<%BANNER%>

Organ Volumes in Pediatric Patients Assessed via Computed Tomography Image Segmentation


PAGE 1

ORGAN VOLUMES IN PEDIATRIC PATIENTS ASSESSED VIA COMPUTER TOMOGRAPHY IMAGE SEGMENTATION By NELLY A. VOLLAND A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2003

PAGE 2

Copyright 2003 by Nelly A. Volland

PAGE 3

To Sophie, Dead on June 5th 1996 from leukemia at the age of 17. She will never get the chance to see what a College student life looks like.

PAGE 4

ACKNOWLEDGMENTS The author would like to thank all the people who have made this project possible. The first thanks go to Dr. Wesley Bolch for the opportunity, the finances, and the computing resources he offered to develop this project. The author would also like to greatly thank Dr. Williams for the time he spent to help find the useful anatomy data needed for the project, check the accuracy of the results, and answer all the anatomy questions. She is also thankful to Dr. Arreola for the time he spent to explain the functioning of computer tomography imaging technique. Special thanks also go to the CT technologists at Shands Teaching Hospital for the kindness, the understanding, and the help they gave during the time spent looking for data in the CT control room. Special thanks also go to Joshua Nipper for the segmentation program he developed, which was very useful during this project and Choonik Lee for all his explanation he could give about this program. Finally, the author would like to thank her family and friends for their support and guidance during this work, which would not have been possible without all of them. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT.........................................................................................................................x CHAPTER 1 INTRODUCTION........................................................................................................1 2 BACKGROUND..........................................................................................................6 Stylized Models............................................................................................................6 Tomographic Models....................................................................................................9 Pediatric Data Sources................................................................................................12 3 MATERIALS AND METHODS...............................................................................23 General Selection and Segmentation..........................................................................23 CT Data Selection................................................................................................23 Segmentation Technique.....................................................................................25 Organ-specific Selection and Segmentation...............................................................28 Brain....................................................................................................................28 Skull.....................................................................................................................30 Lungs...................................................................................................................32 Liver and Kidneys...............................................................................................34 Graphical Analysis......................................................................................................36 4 RESULTS AND DISCUSSION.................................................................................46 Organ Growth Curves.................................................................................................46 Brain....................................................................................................................47 Skull.....................................................................................................................49 Lungs...................................................................................................................51 Liver....................................................................................................................53 Kidneys................................................................................................................54 Comparisons with ICRP Reference Values................................................................54 v

PAGE 6

Characterization of Stylized and Tomographic Models.............................................56 Stylized Models...................................................................................................57 Tomographic Models...........................................................................................58 5 CONCLUSION AND FUTURE WORK...................................................................88 Conclusions.................................................................................................................88 Future Works..............................................................................................................89 LIST OF REFERENCES...................................................................................................94 BIOGRAPHICAL SKETCH...........................................................................................101 vi

PAGE 7

LIST OF TABLES Table page 1-1 Dose equivalent threshold above which different deterministic effects can occur for different organs when irradiated with x-rays........................................................4 2-1 Existing whole-body voxelized models from child to adult....................................16 3-1 Head CT exam characteristics for brain and skull studies.......................................38 3-2 CT exam characteristics for lung study....................................................................40 3-3 CT exam characteristics for liver and kidney studies...............................................41 4-1 Sex, age, brain volume for the 47 children used for the brain study........................63 4-2 Average brain volume during different period of the first year and a half of life and its characteristics...............................................................................................64 4-3 Sex, age, skull volume for the 29 children used for the skull study.........................65 4-4 Average skull volume during different period of the first year and a half of life and its characteristics...............................................................................................66 4-5 Sex, age, lung volume for the 20 children used for the lung study..........................67 4-6 Average lung volume during different period of the first year and a half of life and its characteristics...............................................................................................68 4-7 Sex, age, liver volume for the 11 children used for the liver study.........................69 4-8 Sex, age, kidney volume for the 10 children used for the kidney study..................70 4-9 Mean brain weight per age period............................................................................71 4-10 Mean skull weight per age period............................................................................72 4-11 Mean lung weight per age period.............................................................................73 4-12 Difference between the anthropometric models and the values determined by the growth curves established in this study..............................................................74 vii

PAGE 8

LIST OF FIGURES Figure page 1-1 Photos of necrosis.......................................................................................................5 2-1 Internal view of the stylized model adapted from the MIRD Pamphlet No. 5........17 2-2 External view and cross-sectional view of the Cristy and Ekermans stylized model series..............................................................................................................18 2-3 Head model for the six stylized models as developed by Bouchet et al..................19 2-4 Posterior view of the rectal model for the six stylized model as developed by Marirossian et al.......................................................................................................20 2-5 Models used in radiation dosimetry to assess dose profile received by pediatric patient undergoing diagnostic or interventional examinations................................21 2-6 Height growth rate curve from birth to 18 years of age adapted from the growth in height....................................................................................................................22 3-1 Patient distribution of the brain study by age and sex..............................................42 3-2 Patient distribution of the skull study by age and sex..............................................43 3-3 Patient distribution of the lung study by age and sex...............................................44 3-4 Patient distribution of the liver and kidney studies by age and sex.........................45 4-1 Evolution of the brain volume as a function of age over the first year and a half of life........................................................................................................................76 4-2 Brain volume comparison between CT and MRI volumetry technique...................77 4-3 Evolution of the brain volume as a function of age over the first year and a half of life for male and female.......................................................................................78 4-4 Evolution of the skull volume as a function of age over the first year and a half of life........................................................................................................................79 4-5 3D reconstruction showing the evolution of the skull volume as a function of age over the first year and a half of life....................................................................80 viii

PAGE 9

4-6 Evolution of the total lung volume as a function of age over the first year and a half of life.................................................................................................................81 4-7 Evolution of the left and right lung volume as a function of age over the first year and a half of life................................................................................................82 4-8 Relative difference between left and right lung volume as a function of age over the first year and a half of life..................................................................................83 4-9 Difference between the left and right lung volume for 20 children in their first year and a half of life................................................................................................84 4-10 Difference between the left and right kidney volume difference for 10 children in their first year and a half of life............................................................................85 4-11 Brain mass average evolution for CT-based data and for ICRP data per age period over the first year and a half of life...............................................................86 4-12 Lung mass average evolution for CT-based data and ICRP data per age period over the first year and a half of life..........................................................................87 ix

PAGE 10

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science ORGAN VOLUMES IN PEDIATRIC PATIENTS ASSESSED VIA COMPUTER TOMOGRAPHY IMAGE SEGMENTATION By Nelly A. Volland December, 2003 Chair: Wesley E. Bolch Major Department: Biomedical Engineering Accurate assessments of the radiation absorbed dose to internal organs are needed to assess some level of risk to pediatric patients undergoing radiological procedures. Very young children are considered to be more sensitive to radiation exposures than adults. They can also be uncooperative during the examinations resulting in longer exposures. Furthermore, their smaller size exposes more of their anatomy to the x-ray beam with corresponding increases in their effective dose. Their risk of developing cancer after radiation exposure is also considered to be greater since pediatric patients have a longer post-exposure life expectancy than adult patients. As a consequence, the improvement of quantitative methods for determining organ doses received by pediatric patients undergoing radiological procedures requires the development of anthropometric models that better represent pediatric anatomy. However, very few anthropometric models for children under one year of age have been built to conduct accurate radiation dosimetry calculations. Furthermore, few references are available on organ parameters x

PAGE 11

necessary to properly characterize the few existing models and to develop new models even though organ development is relatively highest during the first year of life. In the present study, a segmentation technique using cross-sectional computed tomography (CT) images was used to establishment organ-specific growth curves for children in their first year and a half of life. In-vivo organ volumes of the brain, the skull, the lungs, the liver and the kidneys were determined based on the number of voxels selected to define the organ and the image voxel characteristics. The organ volume was plotted against the age of the child to determine general organ evolution trends and to provide the needed information to better characterize existing anthropometric models and to help the development of new models in the first year of life. Organ masses were also calculated using reference tissue densities taken from anthropometric model descriptions from the Oak Ridge National Laboratory (ORNL). Some comparisons with the ICRP Publication 23 reference values were made since this report has been used to build anthropometric models for radiation dosimetry. Finally, the characterization of the existing anthropometric models was performed using the growth curves established in this study. The results confirmed that the stylized models were close to a standard population over a large period of time, but not a special individual at a specific age. On the contrary, the tomographic models were shown to be a lot more specific to the individuals they represent and very different from a standard population. The use of cadaver versus living patient to build anthropometric models was also discussed. xi

PAGE 12

CHAPTER 1 INTRODUCTION Medical x-rays represent the main man-made source of radiation and the greatest contribution to radiation doses received by the general public (NCRP 1989). At sufficiently high radiation doses, patients undergoing interventional radiological examinations may be subjected to two different types of effects: deterministic effects and stochastic effects. Deterministic effects mainly occur within the skin, the first and largest organ of the body that the radiation beam will irradiate as shown on the Fig. 1-1. They can also affect eyes or bones during neuroradiology surgery and lungs during cardioradiology surgery because these organs can remain in the primary beam field during the entire intervention. These effects are deterministic because there exists a threshold dose, generally given by the entrance skin dose (ESD), above which they occur and also because they can be quickly detected, even though they may not appear instantaneously. Table 1-1 gives some indications about deterministic effects depending on the dose received by the patient (Geise and O'Dea 1999; Mahesh 2001). For stochastic effects, a threshold dose may not exist since these effects are only related to the probability of developing cancer (carcinogenesis effects) or giving deficient genes to the descendant (hereditary effects). The higher the dose, the more probable the effects are. However, even though the dose received during a single examination represents a low risk, the exposure is cumulative. Determining the total radiation dose 1

PAGE 13

2 for patients undergoing multiple examinations becomes important for further calculations of radiation risk assessment. For very sick patients for whom the diagnosis from imaging examinations can make the difference between life and death, the number of examinations can be numerous, yet necessary. These examinations can be analyzed to minimize the risk by suggesting changes in imaging technique factors to keep the dose received by the patient As Low As Reasonably Achievable (ALARA principle) without losing image quality (NCRP 1990). The Pediatric Organ Dose (POD) project developed at UF is oriented toward the improvement of quantitative methods for determining organ doses received by pediatric patients undergoing radiological procedures. Pediatric patients are especially at risk because they are ten times more sensitive to radiation than adults. They can be uncooperative during the examinations since they do not always understand what is happening. This may force the technologists and radiologists to prolong the procedure, to repeat the procedure, or even both. The generally smaller size of pediatric patients also exposes more of their anatomy to the x-ray beam. This also increases the effective dose received. Since they have a longer post-exposure life expectancy than adults, their risk of developing cancer after radiation exposure is elevated. For all these reasons, trying to understand x-ray examination and to measure radiation dose profile as accurately as possible to assess some level of risk for x-ray imaging for pediatric patients is one of the POD project tasks. To achieve this goal, the development of anthropometric models that better correspond to patients anatomical characteristics is of prime importance.

PAGE 14

3 The development of both physical and computed anthropometric models started when people wanted to quantify radiation hazards and determine necessary protection methods. The stylized models were the first models developed based on simple 3D analytical geometry and on available references of the human body. These models have been used for years for radiation dosimetry purposes. However, tomographic models have become available with the apparition and the improvement of new imaging technologies. These latter models have generated a need for more specific information about the human body to better characterize them. Very few detailed data and references have been published on organ volumes in pediatric patients against which tomographic model data can be compared to, especially absolute organ volumes and masses at a specific age. These comparisons become crucial during the first year of life, since this period corresponds to one of the highest growth rates of the human developmental period. Establishing organ-specific growth curves over this period would allow a better characterization and utilization of the existing stylized and tomographic models, as well as a new database for construction of new models at a given pediatric age. The segmentation of computed tomography (CT) data set can provide the needed information to generate the desired growth curves. The study described hereafter provides the establishment of growth curves over the first twelve months of life for the brain, the skull, the lungs, the liver, and the kidneys, as well as some applications for the existing models.

PAGE 15

4 Table 1-1. Dose equivalent threshold above which different deterministic effects can occur for different organs when irradiated with x-rays. These thresholds are average doses. They can differ more or less from one individual to another. In comparison, the population in the United States of America is exposed on average to the natural background radiation at a dose equivalent rate of approximately 3 mSv/year. Single-Dose Threshold (Sv) Effect 2 Deterministic effects can be detected Temporary erythema Temporary hair loss Cataract 6 Basal skin cells partially damaged Long term erythema Temporary to permanent hair loss Vision impairment 7 Pneumonitis for large lung area exposed 10 Dry desquamation 18 Basal skin cells completely destructed 30 Possible bone changes in young children (Dose per unit exposure in bone is about four times as high as for soft tissue in the range of X-ray used)

PAGE 16

5 A) B) Figure 1-1. Photos of necrosis. A) Skin injury evidencing skin necrosis about 18 to 21 months following fluoroscopy procedures. B) Skin injury about 16 to 21 weeks after the fluoroscopy procedures. A small-ulcerated area is present (Shope 2002).

PAGE 17

CHAPTER 2 BACKGROUND Accurate dose profiles in human anatomy and methods of simulation of radiation transport throughout the body require relatively high definition human body representation. Anatomical modeling for radiation dosimetry was formalized for the first time in 1949 at the Chalk River Conference on Permissible Dose (Chalk River Conference on Permissible Dose 1949). Since this first publication, physical as well as computed models, named as stylized models, have been developed. However, with the development of new computer techniques, tomographic or voxelized models have also started to become available since the late 1980s. This evolution as well as the need for more and more detailed studies has generated a need for more specific information about the human body than the one found in the actual published data and references. Stylized Models The first Standard Man data presented in 1949 (Chalk River Conference on Permissible Dose 1949) described a standardized set of biological parameters that could be used to calculate permissible levels of exposure to workers (adult male). In 1959 after a decade of modifications and improvements the Standard Man was published in the International Commission on Radiation Protections (ICRP) first report (ICRP 1959). Updated in 1975 in the ICRP Publication 23 (ICRP 1975), the Standard Man became the Reference Man (adult male). Biological as well as physiological data were included in this report according to published data primarily derived from autopsy work. Elemental 6

PAGE 18

7 composition for various tissues was also summed up. Some data are available for women and children like masses and organ size, but there are very few. The Medical Internal Radiation Dose (MIRD) Committee of the Society of Nuclear Medicine published its Pamphlet No.5 revised in 1978 (Snyder et al. 1978). In this report, Snyder and his co-workers introduced their latest development done on anthropometric models for internal radiation dosimetry mainly based on the ICRP Publication 23. Simple 3D geometry like ellipsoids, truncated cones, and spheres that better mimic the organ size and shape were used to define this model as shown on Fig. 2-1. Only the male adult was presented. The design of the first pediatric phantoms for 0, 1, 5, 10, and 15 years old started during the same period. Unfortunately, reference data on children were not readily available. As a consequence, the first pediatric models were based on a similitude rule (Hilyer et al. 1972; Poston et al. 1975; Snyder and Cook 1971; Snyder and Ford 1973; Snyder et al. 1976; Warner et al. 1974). This rule transformed the adult phantom into pediatric phantoms using scaling factors in the three dimensions of the space to approximate the dimensions of a child at a given age. This technique produced phantoms that were not realistic pediatric patients. All organs do not have the same evolution throughout the developmental period. The large head, the shorter legs, the thicker trunk, and the large thymus and adrenals of children were completely dismissed in these earlier models. An improvement over these so-called similitude phantoms was done soon after to better represent pediatric patients (Hwang et al. 1976a; Hwang et al. 1976b; Hwang et al. 1976c; Jones et al. 1976). Even though the basic type equations to describe the organs were kept, the organ size, shape, and position were defined according to all the

PAGE 19

8 anatomical references available at the time. In the early 1980s, Cristy and co-workers created a different set of models for dosimetry purposes (Cristy 1980; Cristy and Eckerman 1987). While dimensions of the trunk, the head, and the legs were determined from anthropological data and the organ volumes from ICRP Publication 23 (1975), the similitude rule was used for the organ shape and localization because of a lack of detailed data. Fig 2-2 shows this set of models. When the earliest models have been abandoned, the Cristy and Eckermans models (adult male, 15-year-old child that also represents the adult female, 10-, 5-, 1-year-old children, and newborn) became the unofficial mathematical models for radiation dosimetry. They were upgraded again by Cristy and Eckerman in 1996 (Eckerman et al. 1996). Further improvements have also been made still based on the ICRP Publication 23 recommendations and depending on specific needs. A new head model for the six mathematical models was adopted in 1999 (Bouchet et al. 1999). This new version of the head gives more realistic design of the head and of the neck. Bouchet et al. developed this modified set of heads in response to a demand of more realistic dose calculations for radiopharmaceuticals used in neuroimaging as shown on Fig 2-3. A new rectal model has also been proposed in 1999 (Madirossian et al. 1999) for all the six mathematical models. This revised version of the lower part of the colon includes the rectum, the urinary bladder, and the prostate as shown on Fig 2-4. Dose calculations for radiopharmaceuticals used in brachytherapy and external beam irradiation needed this modified set of the lower part of the colon to be more accurate. A new kidney model has been adopted in 2001 (Bouchet et al. 2003). This revised version of the kidneys describes the specific substructures of this major excretory organ without modifying the

PAGE 20

9 outer dimensions of the existing model. The demand for more detailed local regional doses to specific kidney substructures for beta-particle and low-energy photons irradiation have become of greater and greater importance than only single-region average kidney doses and required the development of this modified set of the kidneys. The stylized models have been developed to represent reference patients following the ICRP publication 23 recommendations and using geometry that approximates as closely as possible the average human body. Small shape variations from these standards are presumed to have small effects on dose calculations and would have necessitated more complex computer programs. Therefore, they were not taken into account to build the models. As a consequence, these models do not represent the patient with a high degree of accuracy. They are hermaphrodite even though a gender specification can still be made depending on the purpose of the study. Finally, they do not represent every age. With the improvement of computational techniques associated to medical imaging, and the appearance on the market of powerful and user friendly image analysis software, some new generation models have been introduced. These new models are called tomographic or voxelized models. Tomographic Models Tomographic or voxelized models can give more realistic representation of human body as they are developed based on the segmentation of medical images of real patients. The complicated three-dimensional shape of patient organs can be better defined in these models than they are in stylized models. This means that the tomographic models would be more adapted than stylized models to better fit patient characteristics at various ages,

PAGE 21

10 especially for pediatric patients who have rapid morphologic changes in a short period of time at different rates and starting at different stages. Several different research groups have already developed some high-resolution tomographic models mainly for adult patients, but also for some pediatric patients, as presented on Table 2-1. They create these models of labeled voxels using three-dimensional data essentially from computed tomography (CT) or magnetic resonance imaging (MRI). In each image data set, each voxel composing the data set is assigned the average characteristic (x-ray attenuation for CT or nuclear magnetic resonance signal for MRI) of the specific region of interest it represents. The smaller the voxel, the more homogeneous its density is and the more accurate the model is, particularly for small organs like ovaries, testes, and adrenals. However, these data do not allow their direct use. The gray level values on the images for different regions of interest overlap each other. Using segmentation techniques allow each voxel to be labeled as belonging to a specific organ, to describe a model, and to define input files suitable for dosimetry transport codes. However, spatial and contrast resolution are most of the time not high enough to perform an entirely automatic segmentation. MR appears to offer better contrast for soft tissues, but the bone tissue definition needed for radiation dosimetry purpose does not appear as good as it does with CT. In addition, MR requires a longer acquisition time and a different reconstruction technique than CT that makes it more sensitive to patient motion artifacts and can severely limit its application to pediatric patients.

PAGE 22

11 The voxelized model was introduced for the first time in the middle of the 1980s (Gibbs et al. 1984; Williams et al. 1986). The two first models created from CT data were published by Zankl et al. in 1988. These models, respectively called BABY and CHILD, were created from a whole-body scan of an 8-week-old female cadaver and a 7-year-old female child who was to undergo a whole-body radiation therapy for leukemia (Zankl et al. 1988). Later, the same team created a tomographic model of a 38-year-old man, called GOLEM, to approximate the Reference Man of the ICRP publication 23 (Zankl and Wittmann 2001). In 1996, Dimbylow created the whole-body voxelized phantom of an adult male adult based on MRI data (Dimbylow 1996). Xu et al. (2000) published the most detailed tomographic voxel model. It represents a male adult referred to as the VIP-man (Xu et al. 2000). It was created from high-resolution photographic images of the Visible Man, published in 1989 by Spitzer and Whitlock (Spitzer and Whitlock 1989). However, the model has been said to be too large to represent the average population. More recently OTOKO (Saito et al. 2001), IRENE and DONNA (Fill et al. 2002) were created from CT data. OTOKO represents an adult Japanese man and IRENE and DONNA two adult females. Two baby voxel phantoms of a 6-day-old female and a 2-month-old male were developed at the University of Florida based on CT data (Nipper et al. 2002). Fig. 2-5 presents the UF-newborn tomographic model with the stylized newborn model. Partial body phantoms have also been developed like the head-torso phantom created in 1994 from CT data and the high-resolution head phantom from MRI images of a healthy volunteer in 1996 (Zubal et al. 1996). ADELAIDE, a 14-year-old trunk model was created (Caon et al. 1999).

PAGE 23

12 If one advantage of the tomographic models is that they are more realistic that stylized models, one disadvantage is that they are more specific. A tomographic model represents the particular characteristics of a specific patient, but not a standard or an average person. As opposed to the stylized model, the tomographic model is built first and then compared to references and also to stylized models. These comparisons help to identify the characteristics of the models in regards to the average population at the same age. However, the characterization of the tomographic models already existing and the construction of other models at various ages in this period can be difficult, because few detailed references are available on organ growth trend especially for young children. Furthermore, over the first year of life, the growth rate is one of the highest of the entire developmental period, as shown on Fig 2-6 for the height (Sinclair and Dangerfield 1978). This growth is not linear. Pediatric Data Sources Whereas patients height, weight, or head circumference can be measured easily and directly, organ volume or organ mass determination is more difficult. Autopsies and fluid displacement studies have been the gold standard methods to establish human body references for many years. Autopsy rates were still above 50% in the 1940s in the US because it was almost the only technique to study anatomy, histology, and disease characterization. However, by 1985 this rate declined to 14% in teaching hospitals and even to less than 9% in non-teaching hospitals (Anguelov 2001). The reasons of this decline are numerous. To perform an autopsy today authorizations from the families are required and most of the time they are difficult to obtain. These studies are also expensive to perform and the infant mortality rate for children in their first year of life has decreased during the last decades. Furthermore, other imaging techniques have

PAGE 24

13 become available that are less evasive to the deceased. X-ray imaging and then ultrasound imaging have been used as the accuracy of autopsy investigations has questioned due to post-mortem tissue changes. Lung tissue changes relatively quickly after death. A progressive but fast fluid accumulation in the lungs postmortem can influence the measurements as mentioned in the ICRP Publication 23. The consequences are a lung volume decrease because of air loss and a lung mass increase because of fluid accumulation postmortem. The liver also contains a high amount of blood whose modification after death can alter its volume determination. Nonetheless, the accuracy of these 2D modality techniques (e.g., radiographs and ultrasound) can also be underlined. The need of simple volume approximations to go from a 2D to a 3D image is obvious. Often organ volumes are overestimated or underestimated (Bakker et al. 1998). However, radiologists still read 2D images for diagnosis and still use published organ measurement tables that give organ characteristic dimensions such as lengths, widths, and depths (Keats and Lusted 1978) measured on these 2D images. Volume measurements, when needed, are usually taken from these 2D dimensions. For these reasons, the stylized models have been built according to the references established with the autopsies and 2D-to-3D calculations, approximations, and interpolations. As mentioned previously, the references used to build these models are indicated in ICRP Publication 23. In this report, the data were taken from different sources. For example, the brain and kidney data were essentially taken from a report of the committee on growth and development of 1933 (Blackman 1933). The data are indicated to be acquired from autopsy work and 2D x-ray imaging. The original source is nevertheless not specified. The outline of physical growth and development published in

PAGE 25

14 1941 (Boyd 1941) as well as from the Introduction to Human Biology and Anatomy for first year Medical Student published in 1952 (Boyd 1952) were used for the lung and liver data. These books summarized Scammons team autopsy work performed in the late 1920s. The improvement of computer technology and the development of cross-sectional and 3D imaging techniques have brought new methods for building anthropometric models that are also useful for organ volume and mass determination. These techniques are able to include the two fundamental conditions (living patients and 3D data) that make measurements of organ volumes and masses closer to reality than prior techniques. Their accuracy and precision have been largely tested and underlined (Disler et al. 1994; Hoe et al. 1997). They started to be used for organ volume and mass determinations in the late 1970s and have proven to be useful for patient diagnosis and therapy (Gault et al. 1988; Hamano and Iwasaki 1990; Heymsfield et al. 1979). However, they have not been used in a systematic manner since high technology skills are required to develop programs that allow quick and easy organ volume and mass determinations. A lot of research has been oriented toward the development of more and more automatic and powerful image segmentation programs. Nonetheless, the different body characteristics and the variety of images techniques usually make studies very specific (Brown et al. 2001; Brown et al. 1999; Rajapakse et al. 1996; Vaidyanathan et al. 1997). Aging and diseases are more often the focus of these studies than young and healthy children (Chapman et al. 1990; Gunduz et al. 2002; Oey et al. 2002; Tsatsanis et al. 2003; Wakabayashi et al. 2002). Furthermore, the entire organ volume is rarely calculated. The recruitment of people who undergo imaging with no underlying disease

PAGE 26

15 is also a complex time, ethical, financial, and practical issue. However, craniosynostosis, which is the early fusion of skull bones in the infancy resulting in an abnormal skull and brain growth, is not uncommon for children. Intracranial volume calculation using 3D imaging has become of more and more interest to study the abnormality. 3D imaging can help to better state intracranial volume evolution, skull surgery, and post-surgery follow-up of the disease. The comparison of the data acquired with a normal growth curve can then be done. Unfortunately, normal growth curves are rare (Abbott et al. 2000; Sgouros et al. 1999). Standard liver volume calculation using CT scans has also appeared to be an accurate and reproducible method that can be beneficial in the assessment of several liver diseases and liver transplantations (Noda et al. 1997; Urata et al. 1995). Standard spleen volume calculations have also been developed due to the increasing number of hypersplenism in children (Watanabe et al. 1997). Overall, few studies have previously performed on whole organ volume calculations using 3D imaging for children in their first year of life, particularly healthy children. Establishing standard organ growth curves for children under one year of age using CT scans segmentation will provide the needed information to better characterize models used in radiation dosimetry.

PAGE 27

16 Table 2-1. Existing whole-body voxelized models from child to adult. Model Gender Age In-plane Resolution (mm) Slice Thickness (mm) # of slices Modality Author UF-Newborn F 6-dy 0.59x0.59 1 485 CT Nipper et al. UF-2-mth M 2-mth 0.49x0.49 1.25 438 CT Nipper et al. BABY F 2-mth 0.85x0.85 4 142 CT Zankl et al. CHILD F 7-yo 1.54x1.54 8 144 CT Zankl et al. IRENE F 32-yo 1.87X1.87 5 ~326 CT Fill et al. VIP-MAN M 38-yo 0.33x0.33 1 1878 Color Photos Xu et al. GOLEM M 38-yo 2.08x2.08 8 220 CT Zankl et al. DONNA F 40-yo 1.87X1.87 10 ~176 CT Fill et al. OTOKO M unknown 0.98x0.98 10 ~170 CT Saito et al. NORMAN M unknown 1.88x1.88 3 ~587 MRI Dimbylow

PAGE 28

17 URINARY BLADDER Figure 2-1. Internal view of the stylized model adapted from the MIRD Pamphlet No. 5.

PAGE 29

18 N ewborn, 1-, 5-, 10-, 15-year-old (adult female), and adult male Figure 2-2. External view and cross-sectional view of the Cristy and Ekermans stylized model series. Adapted from (Cristy 1980; Cristy and Eckerman 1987).

PAGE 30

19 A) B) D) C) E) F) Figure 2-3. Head model for the six stylized models as developed by Bouchet et al. A) The newborn. B) The 1-year-old. C) The 5-year-old. D) The 10-year-old. E) The 15-year-old that also represent the adult female. F) The adult male model.

PAGE 31

20 Figure 2-4. Posterior view of the rectal model for the six stylized model as developed by Marirossian et al. Adapted from (Madirossian et al. 1999) SI, Small Intestine DC, Descending Colon S, Sigmoid Colon R, Rectum UB, Urinary Bladder P, Prostate

PAGE 32

21 A) B) Figure 2-5. Models used in radiation dosimetry to assess dose profile received by pediatric patient undergoing diagnostic or interventional examinations. A) newborn stylized model. B) 6-day-old tomographic model (Nipper et al. 2002).

PAGE 33

22 1 Figure 2-6. Height growth rate curve from birth to 18 years of age adapted from the growth in height (Sinclair and Dangerfield 1978).

PAGE 34

CHAPTER 3 MATERIALS AND METHODS To establish organ-specific standard growth curves for the brain, the skull, the lungs, the liver, and the kidneys, the first task was to collect the in-vivo cross-sectional CT images of children under one year-of-age and to select the more appropriate ones according to the criteria of this study (e.g., normal pediatric anatomy). The second task was the segmentation of the organs to determine their volume and mass in order to build the growth curves. General Selection and Segmentation CT Data Selection The general CT data selection could be divided in three successive steps: (1) the selection based on patient age, (2) the selection based on the patients health condition, and (3) the selection based on imaging sequences themselves. In the first step, all data sets chosen for this study came from CT exams of living patients in their first year and a half of life. The age of the patient was calculated from the date of birth to the examination day. The exams were performed at Shands Teaching Hospital in Gainesville, Florida, between September, 2002 and June, 2003. The data were all acquired on one of four different scanners. The two first scanners were two identical GE light speed 4-slice scanners with a scan field of view (SFOV) of 250 x 250 mm2. Then, two identical Siemens-Somaton sensation 16 scanners replaced them with a SFOV of 500 x 500 mm2. Each image was encoded in a 512 x 512 pixel matrix for both scanners. The protocols used to perform the exams were the ones required for the 23

PAGE 35

24 specific purposes of the study ordered by the physician. As the study was retrospective in nature, no decisions on the imaging techniques were made such as obtaining the thinnest slice thickness achievable for a better axial resolution. Cadavers would have allowed this freedom. However, children cadavers are not as common as adult cadavers and, as mentioned in Chapter 2, family authorization is needed and sometimes difficult to obtain. Furthermore, depending on when the death occurred, tissue transformations may have already taken place and introduced an issue as explained in Chapter 2 for autopsy, especially for lung tissue that changes quickly following death. In addition, no contrast agent is used during the acquisition of the data with cadavers for obvious reasons. This makes the images more difficult to read than the ones acquired on living patients when contrast agents are used. As mentioned in Chapter 2, MR imaging would appear to offer better contrast images for soft tissues, but its longer acquisition time and the different reconstruction technique than CT make it more sensitive to patient motion artifacts and thus severely limits its application to pediatric patients. According to the radiologists, it represents less than 10% of pediatric patients images. For these reasons, only CT images were selected. For the second step of the CT data selection, Dr Jonathan Williams, Chair of Pediatric Radiology at Shands Teaching Hospital, reviewed all the exams selected according to the patients medical history and selected only those scans from patients considered to present normal anatomy at the time of the examination. This means that their health conditions would not drastically alter organ volume or mass. All the studies from patients with diseases or abnormalities that could have resulted in some organ volume or mass errors were simply eliminated.

PAGE 36

25 From there, a further step was to look at the CT sets themselves. CT sets can be composed of several image series. These series do not always have the same number of slices and/or the same reconstruction filters. The different reconstruction filters are usually applied to emphasize the important characteristics needed to be observed in the CT image for the specific study ordered. In this study, the series were chosen in order to make the segmentation easier. The adjacency of the slices within the series and the fully imaging of the organs of interest were also checked. Partial imaging happens commonly as a matter of keeping the radiation dose as low as possible for an exam according to the ALARA principle. Unfortunately, all the data sets that only image partially the organs of interest are useless for this study. Also, all the data sets with overlapping slices or gaps between slices were removed as well, since organ volume determinations done by the segmentation technique were based on the fact that a voxel is associated with one specific region of interest and one specific region is fully defined by a voxel. Segmentation Technique When the data met all the requirements previously described, an in-house segmentation program (Nipper 2002) developed at the University of Florida was used. This program, called CT_Contours, is a Microsoft Windows based program written with Interactive Data Language (IDL) version 5.5 (IDL Version 5.5, Microsoft Windows (Win32x86) 2001). The routine called Ct_Contours was mainly used to perform segmentation and to determine organ volumes. This user-friendly program includes several features such as the conversion of the CT data into a useful format for segmentation, volume calculation, and 3D reconstruction. The conversion consists in transforming the CT images from a DICOM (Digital Communication in Medicine) format into a .jcn format, created by Josh Nipper, to make

PAGE 37

26 the segmentation more practical. This specific function creates a file with an information header followed by the image data set. The information header contains all the useful information such as the data array size and dimension to assess the volume of the region of interests. The image data set becomes a single file rather than the one-file-per-slice technique of the DICOM format. This puts all the slices in the same order of succession and all the parameters that identify the images and the patient are removed. As a consequence, the handling and the segmentation of the data are easier and Health Insurance Portability and Accountability (HIPPA) compliant. The segmentation itself could start when the *.jcn file is loaded in the Ct_Contours routine. Many tools are available with this routine to facilitate the segmentation. As in most medical imaging systems, the window and level features are also available in the Ct_Contours routine. A minimum and a maximum window value can be selected to adjust the contrast of the image to better identify different regions and to conduct automatic segmentation more easily. The values lying between the minimum and the maximum are linearly interpolated between 0 and 255 (8 bits). Among the fifteen different filters available to enhance the image quality, only two were used for this study: the sharpen filter to increase the overall contrast of the regions and sometimes the histogram equalization filter to differentiate tissues in manual segmentation of soft tissue. Different segmentation schemas were run afterwards. Only when the contrast between the region of interest and the surrounding tissues was high enough the segmentation schema could be automatic. If the contrast was not high enough, the segmentation was performed manually. Depending on the area to segment, the semi-automatic thresholding, the pixel and voxel growing segmentation, or the manual freehand

PAGE 38

27 segmentation were used. A Wacom pen tablet Cintiq 15X helped improving the precision and the duration of the segmentation. The manual segmentation was a lot easier and faster using the Cintiq 15X Grip pen direct control on the screen compared to the use of a common computer CRT screen and mouse. When creating the segmentation routine, Nipper (2002) created a color table, named Josh CT Contours, to label each segmented organ region with a specific color tag that would increase its contrast with the surrounding voxels. This same convention was used for simplicity and consistency. This also allowed the differentiation of symmetrical organs such as lungs and kidneys. To obtain the organ of interest volume, an estimator function, available in the routine, calculated the volumes based upon the number of pixels assigned to a specific color within a CT slice and the resolution information found in the *.jcn file header information (slice thickness, pixel matrix size, and image size). A 3D reconstruction routine included in the program, named Xvolume, allowed some reconstructions to be performed. The visualization of the volume slice, the confirmation of the segmentation correctness, and the organ evolution were observed. In radiation dosimetry, the mass of the region of interest is as important as the volume. Doses are calculated in unit of energy per unit of mass, usually mJ per kg (mGy). Furthermore, all the anthropometric models developed until now have been essentially based on the organ mass summarized in the ICRP Publication 23. Using the appropriate density given by Cristy and Eckerman (1980, 1987), the mass of the organs segmented could be determined and then compared to the ICRP data. Report 46 from the International Commission on Radiation Units and Measurements (ICRU 1992) is more detailed and more recent than the Cristy and Eckerman report. However, all the

PAGE 39

28 anthropometric models have been developed using Cristy and Eckermans densities, so they were used in this present study for consistency and comparison purposes. For each organ studied (the brain, the skull, the lungs, the liver, and the kidneys), the selection, the segmentation, and the mass calculation differ slightly from one another as presented further below. Organ-specific Selection and Segmentation Brain Table 3-1 lists all the acquisition parameters for the selected scans for brain segmentation. All the data were taken from Head CT exams. Cases with hydrocephalus, severe hemorrhages, and ventriculo-peritoneal shunt were removed from futher consideration. The group included 47 children from newborns to the age of ~17 months. Thirty-one of the patients were boys (66%). The patients mean age was 4.4 5.14 months. The patient age distribution is presented on Fig. 3-1. Forty-seven percent of the patients were under 3 months of age at the time of the examination. This is explained by the high number of seizure activity early in life. The patients were supine for all the scans. Their heads usually went first through the scanner. The axial mode was the most commonly used acquisition mode. A tilt angle was required sometime depending on the ordered study protocol. The slice thickness varied from 2.5 mm to 5 mm with an average of 3.27 mm. The number of slices per study ranged from 26 to 56 with an average of 37.57 slices per study. The display field of view (DFOV) ranged from 130 x 130 mm2 to 200 x 200 mm2. The in-plan resolution varies from 0.25 mm to 0.40 mm with an average value of 0.32 mm. The voxel volumes were found to be between 0.16 mm3 and 0.73 mm3 with an average of 0.34 mm3. A standard reconstruction filter was frequently used and preferred to the bone

PAGE 40

29 reconstruction filter that was also used on occasion. The reason for this choice is that the bone filter improves significantly the differentiation of the bone content which is not particularly useful for this study of brain volumes. The brain was defined as the group of subregions inside the cranium, which includes the cerebrum, the cerebellum, the diencephalons (constituted of the thalamus and the hypothalamus, the ventricles, and the mesencephalon), the pons, and the medulla oblongata. The spinal cord that began caudal to the medulla oblongata was not included in this definition. The cerebrospinal fluid (CSF) surrounding the brain was not separated from the brain since this differentiation would not be of primary importance for the purpose of this study. This brain definition corresponds to the definition found in the ICRP Publication 23. It has also been called the intracranial volume by some authors (Abbott et al. 2000; Sgouros et al. 1999). The brain is a soft tissue organ surrounded by bone tissues, the skull. The contrast between these two tissues is high enough to perform limited semi-automated segmentations. The 2D and 3D pixel growing segmentation schemas were thus used. The sharpen filter was applied to enhance the brain-skull edges. The window levels were set between 757 and 2187 for entire segmentation of each case. Nevertheless, cranial bones are not completely fused together for most of the children in their first year of life; consequently, the automated growing schemas occasionally expanded the brain tissue segmentation to the surrounding tissues outside the skull. Manual freehand segmentation then had to be applied to remove these superfluous tissues. Some tissues had to be added as well since they are parts of the brain, but were not included during automated

PAGE 41

30 segmentation because their gray level was too different from the tissues surrounding the brain. This usually happened with the fluid-filled ventricles. Cristy and Eckerman (Cristy 1980; Cristy and Eckerman 1987) only defined a generalized soft tissue density of 1.04 g.cm-3, which was used to calculate brain mass. The ICRU Report 46 (ICRU 1992) defines a specific brain density of 1.03 g.cm-3 for a child in his first year of life. This difference is not significant enough to introduce important differences in mass calculation and was not taken into account. Skull Like the brain, head CT exams were chosen for the skull study. All the acquisition parameters of the head CT exams selected for skull volume determination are indicated on Table 3-1 since the exams used for the skull study were the same as the ones selected for the brain study. However, when head CT exams were ordered for brain examination, the skull was often not fully scanned. Therefore, these CT scans were unusable in the case of the skull study. Consequently, there are fewer skull cases than brain cases. Furthermore, some exams were specific cranial exams for which 3D visualization is required afterwards. This demands a high resolution and involves very thin slice thickness and involved image overlaps. As a consequence, these scans also had to be removed from this study. For ease of handling, the case number for the skull volume determination was kept corresponding to the brain case number. The group finally selected for study included 29 children from the newborn to just over 14 months of age. Twenty of the cases (69%) were boys. The patients mean age was 3.51 4.87 months. The patient age distribution is presented on Fig. 3-2. The

PAGE 42

31 reason why there are more cases between 0 and 3 months is still the same as for the brain study since the data are the same. The average image characteristics used in this study are slightly different than those used in the brain study, but this difference is not significant since only few cases have been removed and will not be indicated here. The standard reconstruction filter was preferred over the bone reconstruction filter when possible. The bone reconstruction filter allows a better differentiation of intrabone tissues. The contrast between the bone tissue and the bone marrow is thus enhanced. This would have made the segmentation more difficult than with the standard reconstruction filter since we only want to assess the whole skull volume in this study. The skull normally includes 22 bones: 8 bones of the cranium and 14 bones of the face. The skull is defined here as the bones composing the cranium and the upper face such as parietal, frontal, temporal, occipital, sphenoid, ethmoid, and maxillary bones. The mandible that constitutes the lower face was excluded from the skull definition here so that we remained consistent with the work of Nipper (Nipper et al. 2002). No attempt to dissociate the upper jaw teeth from the maxillary bone was performed. The teeth were thus accounted for in the skull volume. The segmentation of the skull was almost entirely automatic. The contrast was high enough to use the 3D thresholding tools with a threshold between 1110 and 1150 for the pixel value depending on the image quality. The sharpen filter was applied as was done for the brain study with window levels from about 780 to 1604. The sharpen filter helped to enhance the skull-soft tissue edges. Some manual segmentation had to be

PAGE 43

32 performed to remove the bones from the mandible and from the spine as well as medical devices like the CT table, the tubes inside the patient, or the catheters around him/her. Cristy and Eckerman (Cristy 1980; Cristy and Eckerman 1987) defined a bone tissue density of 1.22 g.cm-3 for the newborn and 1.4 g.cm-3 for all other ages throughout the skeleton. These densities represent a homogeneous mixture of bone marrow and the osseous tissues of the bone trabeculae and cortical cortex. A density of 1.61 g.cm-3 has been defined in the ICRU Publication 46 (ICRU 1992) for the cranium of an adult only. This density was thought to be too high for a child in his first year of life. Thus a linear interpolation of the two Cristy and Ekerman bone densities was determined between the newborn and the density applied here for the one-year-old and above. Lungs Table 3-2 lists all the acquisition parameters of the selected scans for lung segmentation. The data were taken from Chest or Chest-Abdomen-Pelvis (CAP) CT exams. Cases presenting expanded abscess, large amount of extra fluid, or severe infections inside the chest were not considered. Every exam for which the lungs were severely collapsed was removed as well because the boundaries between the lungs and the surrounding tissues were usually barely visible and would have led to very large uncertainties. However, exams of slightly collapsed lungs were kept as long as the boundaries of the fully-inflated state were easy to define. The group included 20 children between newborn and ~16 months of age. Fourteen of the patients (70%) were boys. The patients mean age was 7.09 4.72 months. The patient age distribution is presented in Fig. 3-3. There are fewer cases for this lung study than for the brain study because CT is not the primary imaging modality used to identify pediatric chest problems.

PAGE 44

33 The patients were supine and their feet went first through the scanner for almost all these scans, Chest or Chest-Abdomen-Pelvis (CAP) exams. The helical mode was the only acquisition mode used. No scans were performed with a tilt angle. The slice thickness ranged from 1.25 mm to 6 mm with an average of 3.44 mm. The number of slices per study varied from 26 to 101 with an average of 56 slices for chest and CAP studies. CAP studies only have 65, 54 and 49 slices. The display field of view (DFOV) ranged from 132 x 132 mm2 to 250 x 250 mm2. The in-plan resolution varied from 0.26 mm to 0.49 mm with an average of 0.34 mm. The voxel volumes are found between 0.12 mm3 and 0.91 mm3 with an average of 0.42 mm3. Usually the scans were reconstructed with both a standard and a lung reconstruction filter. The standard reconstruction filter was preferred in this study over the lung reconstruction filter. The lung reconstruction filter allows a better differentiation of the contents of the lungs. Consequently, the contrast between the air cavities, the blood vessels, and the bronchial trees that go inside the lung cavities is enhanced. This would have made the segmentation more difficult than with the standard reconstruction filter since we only want to assess the lung cavity volume as a whole in this study. The lungs were defined as the pulmonary cavity including the bronchi and the pulmonary arteries and veins that go inside the lungs. Lungs do not have the same volume during the entire respiratory cycle. Furthermore, a child in his first year of life does not hold his breath when undergoing a CT examination. The CT scan can be taken at any time during the respiration cycle and the lung volume can be influenced accordingly. However, young children breathe either quietly when asleep or quickly but

PAGE 45

34 not deeply when nervous. This negligibly influences the lung volume determination and was not taken into account at all. The segmentation of the lung was semi-automated. The threshold segmentation or the 3D pixel growing segmentation schemes could not be performed since we wanted to separate the left and right lung individually. The 2D region growing and 2D pixel growing segmentation schemes were then selected. The sharpen filter was applied to enhance the lung-soft tissue and lung-rib edges. The window levels were set up between 870 and 1302 for the entire segmentation of each case. However, manual segmentation had to be completed afterwards, either to remove tissues, which were not parts of the lungs, either to add tissues, which was parts of the normal lungs, but not included during automated segmentation. The upper tracheobronchial tree, the lung collapsing, or the amount of extra fluid inside the lung were cases where manual segmentation had to be performed. Lung tissue is defined as a mixture of blood, air, and soft tissue. Only the soft tissue part of lung has a long-term transformation after death. A density of 0.296 g.cm-3 has been indicated by Cristy and Eckerman (1980,1987) for all ages. ICRU Report 46 (ICRU 1992) defined the lung density as 0.26 g.cm-3. Unfortunately, this value is only for adults. The Cristy and Eckermans density was finally chosen since it has been used for all the previous models. Liver and Kidneys Table 3-3 summarizes the acquisition parameters of the CT exams chosen for liver and kidney segmentation. The data were mainly taken from CAP, Abdomen-Pelvis (AP), and abdomen CT scans. One chest study was included because the liver was fully imaged. The cases of expanded abscess or extra fluid, severe infections inside the

PAGE 46

35 abdomen, and the cases of kidney or liver malfunctioning or liver transplant were not considered. The examinations performed without contrast agent were also removed from liver segmentation because the differentiation between the liver and the surrounding tissues was too small to allow accurate segmentation and volume determination. The group included 13 children ranging from newborns to ~16 months. Seven of the patients (54 %) were boys. The patients mean age was 5.65 5.22 months. The patient age distribution is presented on Fig. 3-4. For the same reasons given for the lung study, CT is not the primary imaging modality used to identify pediatric abdomen problems. The patients were supine and their feet went first through the scanner for almost all these exams. Like for the lung study, the helical mode was the only acquisition mode used and no acquisition was performed at a tilt angle. The slice thickness ranged from 2.5 mm to 5 mm with an average of 4.13 mm. The number of slices per study ranged from 36 to 81 with an average of 53.15 slices per study. The display field of view (DFOV) ranged from 118 x 118 mm2 to 223 x 223 mm2. The in-plan resolution varied from 0.23 mm to 0.44 mm with an average of 0.34 mm. The voxel volumes were found between 0.20 mm3 and 0.79 mm3 with an average of 0.49 mm3. Usually the standard and lung reconstruction filters were used for these studies. The standard reconstruction filter was preferred since the lung reconstruction filter was in no way helpful to segment the liver or the kidneys. The liver was defined as a full organ including all interior blood vessels. The gallbladder was excluded even though it was included in most of the autopsy studies (Urata et al. 1995). The liver shape defers from individual to individual as well as from time to time. However, its volume does not largely vary over time.

PAGE 47

36 The kidneys were defined as a full organ including blood vessels, but excluding the urethra. The kidney is one of the only organs in the body that has a very constant shape over time. Its volume is defined as its interior blood supply and the organ parenchyma. The segmentation of the liver was entirely manual. The kidney segmentation was semi-automatic when enough contrast agent was injected into the patient. However, when the contrast was not high enough to use any of the semi-automatic segmentation tools, manual segmentation was performed. The sharpen filter was applied to enhance the contrast, but the histogram equalization filter was used as well to differentiate liver or kidney tissues from the other surrounding soft tissues. The window levels were set between 981 and 1570 for the liver segmentation and between 981 and 1302 for the kidneys segmentation. Cristy and Eckerman (Cristy 1980; Cristy and Eckerman 1987) only defined a soft tissue density of 1.04 g.cm-3, while ICRU Report 46 (ICRU 1992) defined a density of 1.05 g.cm-3 for the liver of a 1-year-old and a density of 1.04 g.cm-3 for a the kidney of a 2-year-old kidney. These differences are not significant and the Cristy and Eckerman density was thus used for both the organs for reasons of consistency discussed previously. Graphical Analysis None of the organs is expected to have a linear growth curve over the first year of life. A graphical data analysis was performed with the software SigmaPlot 2001 (Brannan 2001). The organ volume was plotted against patients age. Best-fit curves were calculated to represent the changes in organ volume that accompany increasing age, creating a model of normal organ growth that is useful for reference purposes. When enough cases were available, a gender differentiation was also performed to emphasize the difference between males and females. While it would have been highly desirable to

PAGE 48

37 obtain patient total mass as the independent parameter for organ volume or mass, this data was not readily available retrospectively, and thus organ volume as a function of patient age is presented only. The number of cases used for this study is low to state any kind of average of organ volume for every month of age. However, to be able to make a comparison with the ICRP Publication 23 data, some age intervals definitions were made. Four or five intervals were defined to match the ICRP Publication 23 intervals: newborn, 0 to 3 months, 3 to 6 months, and either 6 to 12 months and 1 to 2 years, or 6 to 9 months, 9 to 12 months, and 1 to 2 years. As nothing was specified in ICRP Publication 23 about the definition of the newborn, it was decided to define the newborn as babies undergoing CT procedures in their first day-of-life. The 0-to-3-month interval includes babies from 1 day up to 3 month old. The 3-to-6-month interval starts at 3 months and 1 day and ends at 6 month, and so on. All the results will be presented in the next chapter.

PAGE 49

Table 3-1. Head CT exam characteristics for brain and skull studies. Case # for Brain Reconstruction Tilt Angle () # of Slices Slice Thickness (mm) In-Plane Resolution (mm) Voxel Volume (mm3) DFOV (mm) Mode Patient Position Scanner Typea Case # for Skullb B-001 Standard 0 26 5 0.38 0.73 196x196 Axial Head first Supine GE N/A B-002 Standard 0 38 2.5 0.27 0.19 140x140 Axial Head first Supine GE S-002 B-003 Standard 0 40 2.5 0.27 0.19 140x140 Axial Head first Supine GE S-003 B-004 Standard 0 40 2.5 0.26 0.17 135x135 Axial Head first Supine GE N/A B-005 Standard 0 28 3.75 0.29 0.32 150x150 Axial Head first Supine GE S-005 B-006 Standard 0 52 2.5 0.35 0.30 178x178 Axial Head first Supine GE S-006 B-007 Standard 13 40 2.5 0.26 0.17 132x132 Axial Head first Supine GE S-007 B-008 Standard 0 40 2.5 0.28 0.19 142x142 Axial Head first Supine GE S-008 B-009 Standard 0 44 2.5 0.31 0.25 161x161 Axial Head first Supine GE S-009 B-010 Standard 0 42 2.5 0.29 0.22 151x151 Axial Head first Supine GE S-010 B-011 Standard -15 46 2.5 0.28 0.19 142x142 Axial Head first Supine GE S-011 B-012 C30s 0 27 4 0.28 0.31 142x142 Helical Head first Supine Siemens N/A B-013 Standard -17.5 27 5 0.33 0.54 168x168 Axial Head first Supine GE S-013 B-014 Standard 0 50 2.5 0.35 0.31 180x180 Helical Head first Supine GE N/A B-015 Standard 0 32 3.75 0.39 0.57 200x200 Axial Head first Supine GE N/A B-016 Standard 0 40 2.5 0.30 0.23 156x156 Axial Head first Supine GE N/A B-017 Standard 0 56 2.5 0.35 0.31 180x180 Axial Head first Supine GE N/A B-018 Standard 0 50 2.5 0.39 0.38 200x200 Axial Head first Supine GE N/A B-019 Standard 0 44 2.5 0.31 0.24 160x160 Axial Head first Supine GE S-019 B-020 Standard -12 32 3.75 0.31 0.37 160x160 Axial Head first Supine GE N/A B-021 Standard 0 44 2.5 0.30 0.22 152x152 Axial Foot First Supine GE S-021 B-022 Standard 0 40 2.5 0.25 0.16 130x130 Helical Head first Supine GE S-022 B-023 Standard 19 44 2.5 0.31 0.24 160x160 Axial Head first Supine GE N/A B-024 Standard 0 44 2.5 0.28 0.20 144x144 Axial Head first Supine GE N/A B-025 H30s 0 28 4.5 0.30 0.40 152x152 Helical Head first Supine Siemens S-025 B-026 Standard -13 36 3.75 0.33 0.41 169x169 Axial Head first Supine GE S-026 B-027 Standard 0 28 3.75 0.28 0.30 145x145 Axial Head first Supine GE S-027 B-028 Standard -6.5 40 3.75 0.33 0.40 167x167 Axial Head first Supine GE S-028 B-029 H40s 0 41 3 0.35 0.37 180x180 Helical Head first Supine Siemens S-029 38

PAGE 50

Table 3-1. Continued. 39 Case # for Brain Reconstruction Tilt Angle () # of Slices Slice Thickness (mm) In-Plane Resolution (mm) Voxel Volume (mm3) DFOV (mm) Mode Patient Position Scanner Typea Case # for Skullb B-030 H40s 0 28 4.5 0.29 0.38 148x148 Axial Head first Supine Siemens S-030 B-031 H40s 0 36 3 0.33 0.33 170x170 Axial Head first Supine Siemens S-031 B-032 Standard 0 36 2.5 0.27 0.19 140x140 Helical Head first Supine GE S-032 B-033 H30s 0 28 4.5 0.30 0.41 155x155 Axial Head first Supine Siemens S-033 B-034 C30s 0 28 4.5 0.35 0.56 180x180 Axial Head first Supine Siemens S-034 B-035 Standard 0 36 3.75 0.39 0.57 200x200 Axial Head first Supine GE N/A B-036 H30s 0 28 4.5 0.35 0.56 180x180 Axial Head first Supine Siemens N/A B-037 H30s 0 32 4.5 0.35 0.56 180x180 Axial Head first Supine Siemens S-037 B-038 H30s 0 42 3 0.29 0.26 150x150 Axial Head first Supine Siemens S-038 B-039 C30s 0 47 3 0.34 0.35 176x176 Helical Head first Supine Siemens N/A B-040 H40s 0 30 3 0.27 0.22 140x140 Axial Head first Supine Siemens N/A B-041 C30s 0 36 3 0.25 0.19 130x130 Axial Head first Supine Siemens S-041 B-042 C30s 0.5 32 4.5 0.35 0.56 180x180 Axial Head first Supine Siemens S-042 B-043 H40s 0 32 4.5 0.39 0.69 200x200 Axial Head first Supine Siemens S-043 B-044 H40s 0 42 3 0.40 0.48 204x204 Axial Head first Supine Siemens N/A B-045 H40s 0 36 3 0.26 0.21 134x134 Axial Head first Supine Siemens N/A B-046 H40s 0 42 3 0.38 0.43 194x194 Axial Head first Supine Siemens N/A B-047 H40s 0 36 3 0.34 0.34 172x172 Axial Head first Supine Siemens N/A a GE, GE light Speed 4-slice scanner; Siemens, Siemens-Somaton sensation 16 scanner b N/A, CT head exams cannot be used for skull volume calculation

PAGE 51

Table 3-2. CT exam characteristics for lung study. Case # for Lung Study typea Reconstruction # of Slices Slice Thickness(mm) In-Plane Resolution (mm) Voxel Volume (mm3) DFOV (mm) Mode Patient Position Scanner Typeb LU-001 CAP B30f 49 5 0.30 0.45 154x154 Helical Foot First Supine Siemens LU-002 Chest B30f 52 2 0.27 0.15 140x140 Helical Head First Supine Siemens LU-003 Chest B30f 26 5 0.26 0.34 134x134 Helical Foot First Supine Siemens LU-004 CAP B30f 54 5 0.31 0.48 158x158 Helical Foot First Supine Siemens LU-005 Chest Standard 101 1.25 0.31 0.12 160x160 Helical Foot First Supine GE LU-006 Chest Standard 73 2.5 0.49 0.60 250x250 Helical Foot First Supine GE LU-007 Chest Standard 57 2.5 0.36 0.33 186x186 Helical Foot First Supine GE LU-008 Chest B30f 79 2 0.32 0.21 166x166 Helical Foot First Supine Siemens LU-009 Chest Standard 45 3.75 0.45 0.74 228x228 Helical Foot First Supine GE LU-010 Chest Standard 63 2.5 0.39 0.39 201x201 Helical Foot First Supine GE LU-011 Chest Standard 41 2.5 0.40 0.39 203x203 Helical Foot First Supine GE LU-012 Chest B30f 56 3 0.33 0.33 170x170 Helical Foot First Supine Siemens LU-013 Chest Standard 49 3.75 0.41 0.64 211x211 Helical Foot First Supine GE LU-014 CAP B30f 65 5 0.32 0.51 164x164 Helical Foot First Supine Siemens LU-015 Chest B30f 34 6 0.39 0.91 199x199 Helical Foot First Supine Siemens LU-016 Chest B30f 68 2 0.26 0.13 132x132 Helical Head First Supine Siemens LU-017 Chest B30f 83 2 0.31 0.20 161x161 Helical Foot First Supine Siemens LU-018 Chest B30f 51 3 0.35 0.37 180x180 Helical Foot First Supine Siemens LU-019 Chest B30f 36 5 0.35 0.62 180x180 Helical Foot First Supine Siemens LU-020 Chest B30f 35 5 0.30 0.45 154x154 Helical Foot First Supine Siemens 40 a CAP, chest-Abdomen-Pelvis Study b GE, GE light Speed 4-slice scanner; Siemens, Siemens-Somaton sensation 16 scanner

PAGE 52

Table 3-3. CT exam characteristics for liver and kidney studies. 41 Case # for Liver Study Typea Reconstruction # of Slices Slice Thickness (mm) In-Plane Resolution (mm) Voxel Volume (mm3) DFOV (mm) Mode Patient Position Scanner Typeb Case # for Kidneyc N/A AP Standard 81 2.5 0.35 0.31 180x180 Helical Foot First Supine GE K-001 LI002 Abdomen Standard 53 2.5 0.35 0.31 180x180 Helical Foot First Supine GE N/A LI003 AP Standard 69 3.75 0.44 0.71 223x223 Helical Foot First Supine GE K-003 LI004 AP B30f 44 5 0.40 0.79 204x204 Helical Head First Supine Siemens K-004 LI005 AP Standard 42 3.75 0.23 0.20 118x118 Helical Foot First Supine GE K-005 LI006 AP Standard 68 3.75 0.43 0.70 221x221 Helical Foot First Supine GE K-006 LI007 AP Standard 40 2.5 0.31 0.24 160x160 Helical Foot First Supine GE K-007 LI008 CAP B30f 49 5 0.30 0.45 154x154 Helical Foot First Supine Siemens K-008 LI009 CAP B30f 54 5 0.31 0.48 158x158 Helical Foot First Supine Siemens K-009 LI010 CAP B30f 65 5 0.32 0.51 164x164 Helical Foot First Supine Siemens N/A LI011 AP B30f 40 5 0.30 0.44 152x152 Helical Foot First Supine Siemens K-011 N/A AP B30f 50 5 0.35 0.62 180x180 Helical Foot First Supine Siemens K-012 LI013 Chest B30f 36 5 0.35 0.62 180x180 Helical Foot First Supine Siemens N/A a AP, Abdomen-Pelvis Study; CAP, chest-Abdomen-Pelvis Study b GE, GE light Speed 4-slice scanner; Siemens, Siemens-Somaton sensation 16 scanner c N/A, CT exams cannot be used for liver or kidney volume calculation

PAGE 53

42 024681012141600-11-22-33-44-55-66-77-88-99-1010-1111-1212-1313-1414-1515-1616-17Age (months)Number of Patients Female Male Figure 3-1. Patient distribution of the brain study by age and sex.

PAGE 54

43 02468101200-11-22-33-44-55-66-77-88-99-1010-1111-1212-1313-1414-1515-1616-17Age (Months)Number of Patients Female Male Figure 3-2. Patient distribution of the skull study by age and sex.

PAGE 55

44 0123400-11-22-33-44-55-66-77-88-99-1010-1111-1212-1313-1414-1515-1616-17Age (months)Number of Patients Female Male Figure 3-3. Patient distribution of the lung study by age and sex.

PAGE 56

45 0123400-11-22-33-44-55-66-77-88-99-1010-1111-1212-1313-1414-1515-1616-17Age (months)Number of Patients Female Male Figure 3-4. Patient distribution of the liver and kidney studies by age and sex.

PAGE 57

46 CHAPTER 4 RESULTS AND DISCUSSION The results and discussion will be divided in three parts: (1) the establishment of the organ-specific growth curves, (2) the comparison with refe rences and already published data if any, and (3) the character ization of the stylized and tomographic models. Organ Growth Curves In this first step of the establishment of the organ-specific growth curves, the general trend of organ evolution was determined and will be detailed later in this chapter for the brain, the skull, and the lungs. Th ree-parameter hyperbola and three-parameter power equations were found to be the best-fit curve empirical equations for the different organ growth curves presented separately he reafter. The three-parameter hyperbola is described by the equation: 0aAGE VOLV bAGE (4-1) In Eq. (4-1) VOL represents the organ volume in cm3 at a given age, V0 is the supposed organ volume at birth, a and b are two coefficients that will characterize the growth curve, and AGE is the age of the child in months. The three-parameter power equation is described by: 0 dVOLVcAGE (4-2)

PAGE 58

47 In Eq. (4-2) VOL represents the organ volume in cm3 at a given age, V0 is the supposed organ volume at birth, c and d are two coefficients that will characterize the growth curve, and AGE the age of the child in months. When it was possible, a gender distincti on was made. This data differentiation usually demonstrates the differe nce between girls and boys in a manner similar to that for height, weight, and head circumference (Clinical Growth Charts 2001). All the specificities of each organ studi ed are detailed hereafter. Brain Table 4-1 lists the characteristics of each patient selected for the brain study including gender, age, and brai n volume. Forty-seven brains have been segmented from newborn to ~17 months of age. The smallest brain is only 258.85 cm3 for a 0.3 monthold child and the largest is 1232.03 cm3 for 14.29 month-old. There are important differences between individuals even at the sa me age. Fig. 4-1 was created from the data of Table 4-1. The brain volume is plotted agains t the age of the patient. This represents a model of normal brain growth over the first ye ar-of-life regardless the gender. The data spread observed in the first months of life can reflect the difference between babies born prematurely or at normal gestation. Unfort unately, limited access to the medical records did not allow identification of those born prematurely. None theless, premature birth does not make any difference after a couple of months (Ounsted et al. 1982). The principal differences essentially come from family ch aracteristics, like pare nts height, and the severity of birth defects and il lnesses. However, patients in serious conditions have been withdrawn from this study. Only patients w ith nominally healthy condition remain. The main factors of difference in this study are finally the premature bi rth in the first few months of life and the parents morphometric characteristics.

PAGE 59

48 The general trend is a rapid growth in the first weeks of life and then the rate slows with increasing age. The most appropriate e quation to fit the curve has been found to be described by the three-parameter hyperbola: 1092.77 374.04 8.762AGE VOL AGE (4-3) The coefficient of determination (R2) is 0.84. Sgouros et al. (1999) did a similar study for patients from age 7 days to 15 year s using magnetic resonance imaging. When comparing the brain volume (calle d intracranial volume in Sgouros et al. (1999) study) from the patients under one year of age (13 cas es) of this study with the brain volume of our study, the results are similar to those shown in Fig. 4-2. Abbott et al. (2000) did not specify their results as only graphical data we re published. Furthermore, they determined the age of the children from conception to th e day of the examination. Therefore, the comparison would not have been as relevant. Fig 4-3 shows the difference in normal brain development between males and females. Male brain volumes best-fit curve is constantly above the female fit curve. This agrees with head circumference charts, in which a similar difference between the two genders has been established (Clinical Growth Charts 200 1). The most appropriate equations to fit these curves have also been found to be th ree-parameter hyperbolas. For females, the equation has been defined as: 912.46 288.27 4.67AGE VOL AGE (4-4) The coefficient of determination (R2) is 0.77. For males, the equation has been defined as: 1396.8 394.2 13.6AGE VOL AGE (4-5)

PAGE 60

49 The coefficient of determination (R2) is 0.90. In most of the published data, the curves for males and females differ from one a nother but the evolution is parallel (Abbott et al. 2000; Sgouros et al. 1999). Here this is not the cas e. A possible explanation is the small number of female data (only 15 cases). Table 4-2 summarizes the average volumes ov er five different periods of time in the first year-of-life. It al so indicates the minimum and the maximum brain volumes for each period as well as the standard deviation an d the percentage difference relative to the mean (column 2) between the maximum and minimum volumes. For a newborn, the mean brain volume is 406.83 cm3. The smallest brain volume is 331.57 cm3 and the largest is 518.92 cm3. This volume reaches 451.08 cm3 during the first three months of life. It increases to 697.51 cm3 on average between the ages of 3 and 6 months to approach 940 cm3 between 6 and 12 months of age. This average attains 1010.24 cm3 between 1 and 2 years of age. This repr esents a 250 % brain volume increase over the first year-of-life. Skull Table 4-3 lists each patient selected for the skull study including gender, age, and skull volume. There are large differences be tween individuals even at the same age. Twenty-nine skulls have been segmented from the newborn to ~14 months. The smallest skull is only 58.17 cm3 for a 0.3 month-old and the largest is 344.07 cm3 for a 14.29 month-old. These results are in correla tion with the corresponding results for brain volume. Fig. 4-4 was created from the data of Table 4-3. The skull volume is plotted against the age of the patient. This repres ents a model of normal skull growth over the first year-of-life regardless the gender.

PAGE 61

50 The general trend is the same as the brai n volume growth, since the brain can also extend if the skull extends. The most appropr iate equation to fit th e curve has been found to be a three-parameter hyperbola described by: 441.24 81.74 14.38AGE VOL AGE (4-6) Equation (4-6) is represente d on Fig. 4-4. The coe fficient of determination (R2) is 0.91. Unfortunately there are no published data with which to make comparisons. Medically, the skull volume is of little clin ical interest. The most important skull characteristics monitored are usually th e intracranial dimensions (Waitzman et al. 1992a; Waitzman et al. 1992b) to sometimes determine the intracranial volume. These dimensions are critical for most of the cran iofacial surgeries like for craniosynostosis. The volume of the skull bone (Buda et al. 1975) is rarely measured. The difference in normal skull development between males and females is not presented here, because the number of cases for female children is not high enough (9) to allow any comparison with the male children (20). Table 4-4 summarizes the average volumes ov er five different periods of time in the first year of life. It also indicates the minimum and the maximum skull volume for each period as well as the standard deviation an d the percentage difference relative to the mean (column 2) between the maximum and mi nimum volume. For a newborn, the mean skull volume is 87.56 cm3. The smallest skull volume is 78.79 cm3 and the largest is 101.43 cm3. This volume reaches 102.89 cm3 during the first three months of life. It increases to 154.89 cm3 on average between the age of 3 and 6 months to attain 232.6 cm3 between 6 and 12 months of age. This average approaches 300 cm3 between 1 and 2 years of age. This represen ts a 340 % skull volume increas e over the first year-of-life.

PAGE 62

51 If the skull volume may not have an im portant interest out of the radiation dosimetry field, the segmentation allowed 3D reconstruction that shows another view of the skull evolution. The growth of the di fferent bones of the skull until they all fused together is well imaged with this reconstruction. If the fontanels are large at birth, they almost disappear within the first few months of life as shown on Fig. 4-5. Lungs Table 4-5 lists each patient selected for the lung study including gender, age, total lung volume, left and right lung volume, and the difference between the left and right lung volume. The right lung is compared to th e left lung. Twenty pairs of lungs have been segmented from the newborn to ~16 mont hs. The smallest pair of lungs is only 163.12 cm3 for a 0.23-month-old and the largest is 621.13 cm3 for 13.6-month-old. There are large differences between individuals even at the same age. Fig. 4-6 was created from the data of Table 4-5. The total l ung volume is plotted against the age of the patient. This represents a model of nor mal lung growth over the first year-of-life regardless the gender. The general trend is different from the brain and the skull studies. The lungs display a nearly constant growth rate over the first year -of-life. The most appropriate equation to fit the curve has been found to be a three-parameter power equation described below and re presented on the Fig. 4-6. 0.68167.4248.62 VOLAGE (4-7) The coefficient of determination (R2) is 0.55. Unfortunately there is no published data with which to make comparisons for our study. The only publ ished data for lung volume calculations are essentially performe d for elderly people with lung cancer or chronic diseases.

PAGE 63

52 The difference in normal lung development between males and females is not presented in this study because there are no t enough cases for both genders (6 female cases and 14 male cases). However, the left and right lung devel opment is indicated in Fig. 4-7, as well as the difference between th eir volumes in Fig. 48. The left lung is generally smaller than the right lung. The difference between both l ungs can greatly vary between individuals even in the same age categ ory. The most approp riate equations to fit these curves have also been found to be three-parameter power equations. For the left lung, the equation ha s been defined as: 0.6267.9128.22 VOLAGE (4-8) The coefficient of determination (R2) is 0.53. For the righ t lung, the equation has been defined as: 0.7499.0521 VOLAGE (4-9) The coefficient of determination (R2) is 0.55. Nonetheless, no difference in evolution over the time can be set as shown in Fig. 4-8 and in Fi g. 4-9. The largest difference between the left and the right lung is 37 % for a 1.97-month-old and the smallest is 1 % for a 8.71-month-old. The overall average difference is 16 % and gives an average left lung volume 45.6 % of the tota l lung volume (TLV). This agrees with the left lung accounting for between 45.5% a nd 49.1% of the TLV in different studies already performed on adults using different methods (Denison et al. 1986; Horsefield 1967; Pierce et al. 1980). The constant average diffe rence between left and right lung throughout life is emphasized in this study. In one case the right lung is smaller than th e left lung. This situation is referred to as a volume loss. This can be occurring due to a normal congenital variation, a

PAGE 64

53 pneumonia, which is a pulmonary infection, or even an atelectasis, a bronchi collapsing due to mucus. This is not uncommon. In this particular case, it was not considered severe enough to be removed from our study. The case where the left lung is significantly smaller than the right lung can also reflect the volume loss problem that usually disappears over time. Table 4-6 summarizes the average volumes over five different periods of time in the first year-of-life. It also indicates for each period as well as the standard deviation and the percentage difference relative to the mean (column 2) between the maximum and minimum volume. The total lung volume is 200 cm3 on average during the first three months-of-life. It increases to 337.28 cm3 between the age of 3 and 6 months to attain 390.5 cm3 between 6 and 9 months of age. This average approaches 357 cm3 between 9 and 12 months of age to reach 516.38 cm3 between 1 and 2 years of age. The average total lung volume for the 9-to-12-month-old period is smaller than the 6-to-9-month-old period. This may be due to the small number of cases for both periods. The cases may not truly represent an average population even though the volume difference over these periods is small. The volume increases by 260 % over the first year-of-life. Liver Table 4-7 lists each patient selected for the liver study including gender, age, and the liver volume. Eleven livers were segmented from the newborn to ~16 months. The smallest liver is only 148.8 cm3 for a 0.23 month-old and the biggest is 489.33 cm3 for 12.39 month-old. There are large differences between individuals even at the same age. However, there are not enough cases to conduct a complete study like the brain, the skull, or the lung studies presented previously.

PAGE 65

54 Kidneys Table 4-8 lists each patient selected for the kidney study including gender, age, total kidney volume, left and right kidney volume, and the difference between the left and right lung volume. Ten pairs of kidneys were segmented from the newborn to ~12 months. The smallest pair of kidneys is only 21.46 cm3 for a 0.03 month-old and the largest is 95.52 cm3 for a 12.39 month-old. There are large differences between individuals even at the same age. In general, the left kidney is slightly larger than the right kidney. This statement agrees with the ICRP Publication 23, although the quantification of this difference is not indicated. In this study, the largest difference is 20% for the 12.39-month-old and the smallest is 1 % for a 2.17-month-old. The overall average difference is 7%. Fig. 4-10 represents this difference. Brown et al. (2001) mentioned a 4% different in their study of 8 pediatric kidneys. In two cases the right kidney is larger than the left kidney. There is no explanation for this fact. However, this study has too few cases to go further than giving the volume and mass results. Any statement would not be speculative. Comparisons with ICRP Reference Values In the second part of the study, some comparisons with the ICRP Publication 23 were performed because this report has always been the reference to build anthropometric models for radiation dosimetry. In this report the main organ references are organ mass as differentiated by gender. As a consequence, before making some comparisons with the ICRP Publication 23 values, organ masses were calculated using the CT-based organ volumes and the Cristy and Eckermans densities indicated in Chapter 3 (Cristy 1980; Cristy and Eckerman 1987). The average over both genders was also calculated for the

PAGE 66

55 ICRP Publication 23 data to allow the comparison between CT-based and ICRP-report organ masses. The CT-based values were too few to allow a gender comparison. Table 4-9 lists the CT-based brain masses and the ICRP 23 brain masses for five period of growth. Fig. 4-11 represents graphically the difference between the two data sets. The CT-based brain mass averages are larger than the ICRP averages one for each period. Table 4-9 also indicates the percentage difference between the CT-based brain masses and the ICRP 23 brain masses for each period. These differences range from 10.7% for the 0-to-3-month-old period to 28% for the 3-to-6-month-old period. The ICRP publication 23 does not give any specific skull masses throughout the life time. The total weight for the skeleton is given on a graph. Specific bones are characterized by a percentage of this total skeleton weight for newborn and adult only. However, the average masses of the skull over 5 periods from birth to 12 months were calculated. Table 4-10 summarizes these CT-based skull masses. Table 4-11 summarizes the CT-based total lung masses and the ICRP total lung masses for five periods. Fig. 4-12 represents graphically the difference between the two data sets. The CT-based lung mass averages are smaller than the ICRP averages one for each period. Table 4-11 also indicates the percentage difference between the CT-based total lung masses and the ICRP 23 total lung volume for each period. These differences go from -26% for the 9-to-12-month-old period to 4% for the 3-to-6-month-old period. No left and right lung masses are specified in the ICRP Publication 23. Table 4-7 summarizes the CT-based liver masses and Table 4-8 the CT-based kidney masses. However, for the same reason enounced previously no comparison can be made. Too few cases are available.

PAGE 67

56 Over all for mass comparison, a debatable issue can be enounced. The ICRP publication 23 gathers a lot more cases per period than this present study. However, these differences can be attributed to a combination of factors such as (1) overall increases in individual size and stature of the past several decades, (2) problems with tissue swelling or shrinkage post mortem, and (3) differences in techniques used for brain volume and/or mass determination like it has already been stated in some publication (Abbott et al. 2000; Sgouros et al. 1999). Characterization of Stylized and Tomographic Models The third and last part of the study was the characterization of the existing anthropomorphic models using the growth curves established earlier in this chapter. Only organ volumes were compared since organ mass comparisons would not have brought different results. The same densities were used in the development of the anthropometric models and the growth curves. The differences between the model organ volume and the growth curves values were stated per organ and for the entire model. For each organ the relative difference between the model and the growth curve values was calculated. For the entire model, the difference took into account the average of the absolute differences for the brain, the skull and the TVL. The equation was defined by: organaveragemodeln (0-10) In Eq. (4-10) represents the average difference between the model and the growth curve established above. averagemodel organ is the relative difference between the model

PAGE 68

57 and the growth curve values. n is equal to the number of organs used to calculate the average, n=3 in this study. Stylized Models To build the anthropomorphic stylized model Cristy and Eckerman followed the references stated in the ICRP Publication 23. The brain, the skull, and the lung volumes they used are summarized in Table 4-12 for the newborn model and for the one-year-old model. When Bouchet et al. (1999) made a new head model, the ICRP Publication 23 was still the preferred reference source for a manner of consistency. However, the brain volume has been increased by 10% compared to the Cristy and Eckermans brain model for the newborn, giving a new volume of 371.9 cm3. This volume became 1.4% smaller for the one-year-old model, going from 850 cm3 to 838.4 cm3. Even with an augmented brain volume, Bouchet et al. model is still small compared to the average values obtained in this study (406.83 cm3). However, this new volume (371.9 cm3) is only 1% smaller than the newborn value determined by Eq. (4-1) defined previously in this study (374.04 cm3). Unfortunately, the new one-year-old brain volume is still very low. The volume defined by Eq. (4-1) above is 1005.7 cm3 for a 12-month-old child when for the one-year-old stylized model the brain volume is only 838.4 cm3 using the new head model of Bouchet et al.. Cristy and Eckermans skull model includes the mandible because the facial bones were undifferentiated, unlike Bouchet et al. stylized head model and the skull defined in this study. The skull volume has been increased by 50 % in Bouchet et al. (1999) new head model compared to the Cristy and Ekermans skull model for the newborn, giving a new volume of 83.5 cm3. This volume became 56 % bigger for the one-year-old model, going from 161.8 cm3 to 252.8 cm3. The newborn skull volume from Bouchet et al.

PAGE 69

58 appears only 5% smaller compared to the average values obtained in this study (87.56 cm3). However, this new volume (83.5 cm3) is less than 1% bigger than the newborn value that is determined by Eq. (4-6) defined previously in this study (81.74 cm3). The new one-year-old skull volume is also within 0.5% difference with the value defined by Eq. (4-6). The volume defined by Eq. (4-6) above is 282.46 cm3 for a 12-month-old child when for the one-year-old stylized model the skull volume is 252.8 cm3 using the new head model from Bouchet et al.. The lung volume for the newborn stylized model is 171 cm3, the left lung volume is 79.1 cm3 and the right lung volume is 91.9 cm3. This represents a 14% difference between the left and right lung volume. The total volume is about 2% bigger than the one given by Eq. (4-7) established previously (167.42 cm3). However, the left lung volume is 16% bigger and the right one is 7% smaller than the results provided by Eq. (4-8) and Eq. (4-9) (67.91 cm3 for the left lung and 99.05 cm3 for the right one). The lung volume for the 1-year-old stylized model is 484 cm3, the left lung volume is 225 cm3 and the right one is 259 cm3. This represents a 13% difference between the left and right lung volume. The total volume as well as the left and right lung volume is about 12% bigger than the one given by Eqs. (4-7), (4-8), and (4-9) (430.85 cm3, 199.63 cm3, and 231.12 cm3, respectively). Tomographic Models The brain, the skull, and the lung volumes found for the UF-newborn (6-day-old child) female, the UF-2-month, and the UF-9-month male tomographic model are also summarized in Table 4-12 (Nipper et al. 2002).

PAGE 70

59 The UF-newborn brain is smaller than the stylized model and 22 % smaller than the value defined by Eq. (4-1) (398.43 cm3). However, if Eq. (4-2) is used, the brain volume for a 6-day-old female should approach 325.74 cm3, that only represents a 5% difference. Being aware that the female brain volume evolution in function of the age over the first year of life was established with a small number of cases compared to the male and the undifferentiated brain volume growth curve, this result can still be debatable. Its skull has a volume of 97.34 cm3 (Nipper et al. 2002). This value is 10 % bigger than the stylized model and 9 % bigger than the value defined by Eq. (4-6) (87.79 cm3). Its total lung with a volume of 122.83 cm3, the left lung reaches 56.5 cm3, when the right one reaches 66.33 cm3 to make a 15% difference. This pair of lungs is not the cadaver used to build the model. It comes from a living patient who had approximately the same characteristics than the cadaver who had largely collapsed lungs. It appears smaller than the stylized model lungs and about 30% below the curve defined by the Eqs. (4-7), (4-8), and (4-9) (183.69 cm3, 78.31 cm3, and 105.43 cm3, respectively). The UF-2-month male tomographic model has been found to have a brain volume of 535.97 cm3 (Nipper et al. 2002). Eq. (4-1) gives a brain volume of 577.16 cm3 for a 2-month-old child and Eq. (4-3) 573.28 cm3. This represents only about 7 % difference for both equations. Its skull has a volume of 166.98 cm3 (Nipper et al. 2002). Eq. (4-6) for a 2-month-old child gives a skull volume of 135.62 cm3. This represents about 23% difference. Its total lung volume was determined at 156.11 cm3. The left lung was 74.53 cm3 and the right 80 cm3. This represents a 7% difference between the left and right lung. Eq. (4-7) gives a total lung volume of 245.32 cm3 for a 2-month-old child, Eq. (4-8) gives a left lung volume of 111.28 cm3, Eq. (4-9) 134.12 cm3 for the right lung

PAGE 71

60 volume. This represents a difference over 32% for the three equations. However, the CT data used to build this model came from a 6-month-old male cadaver. Eq. (4-1) gives an 818.25-cm3 brain volume for a 6-month-old child and Eq. (4-3) 821.79 cm3. This represents about 1.5 times the actual child brain volume determined by segmentation. Eq. (4-6) gives a 211.64-cm3 skull volume for a 6-month-old child. This represents more than 1.25 times the actual child skull volume determined by segmentation. Eq. (4-7) gives a 331.84-cm3 total lung volume for a 6-month-old child, Eq. (4-8) gives 153.92 cm3 for the left lung volume and Eq. (4-9) 178.13 cm3 for the right lung volume. This represents about 2 times the actual child lung volume determined by segmentation. This child was born highly premature and was very sick. His height and weight was checked and better corresponded to an average 2-month-old child than to a 6-month-old child according to the growth charts (Clinical Growth Charts 2001). A 9-month-old male model has recently been built here at UF. Its brain volume approaches 948 cm3. However, Eq. (4-1) gives a 927.8-cm3 brain volume for a 9-month-old child when Eq. (4-3) gives 950.45 cm3. The difference is then 3% for the general trend and 1% for the male trend. His skull volume approaches 323 cm3. Eq. (4-6) gives a 251.59-cm3 brain volume for a 9-month-old child. The difference falls within 22 %. Its total lung volume approaches 335 cm3, its left lung 152 cm3, and its right lung 183 cm3. Eq. (4-7) gives a 384.04-cm3 total lung volume for a 9-month-old child when Eq. (4-8) gives 178.11 cm3 for the left lung volume and Eq. (4-9) gives 205.8 cm3 for the right lung volume. The difference is 11% for the right lung and 15% for the left lung.

PAGE 72

61 From the results presented in Table 4-12 stylized and tomographic models were compared to the curves established in this study. The stylized models with Bouchet et al. head model have been observed to nicely represent the average population in terms of volume even though these models are very simplistic. The tomographic models trend to behave differently. The UF-newborn model is 22% smaller than average over the organs used. The lungs were imported from a living patient. They are particularly small, even though their match with the body is visually appropriate. The whole-body model is simply small even if the patient had an average weight for a newborn (3.84 kg) according to the weight CDC growth chart (Clinical Growth Charts 2001). Post mortem changes may already have taken place. The UF-2-month model was definitively not equivalent to a 6-month-old average baby. The age scaling done according to the total weight and height of the patient appears appropriate, although the model still has small lungs. The brain volume appears in the range of a 2-month-old child, but for the skull and the lungs, it does not appear appropriate. The lungs are smaller than the newborn lung volume determined by the growth curve (Eq. (4-7)). The skull would match a 4-month-old skull according to the growth curve (Eq. (4-6)). With more organ comparison it would able to have more detailed conclusions. However, it can be stated that this patient was sick and weight and height can be a better indicator of child development than age. The 9-month-old model is only about 14% different on average for the whole-body. The differences with the growth curves appear smaller than the two other tomographic models. One explanation may be the fact that this model has been developed from data

PAGE 73

62 of a living patient and not a cadaver like the two other models. However, the reasons for performing the full CT scan on this particular patient were not determined.

PAGE 74

63 Table 4-1. Sex, age, brain volume for the 47 children used for the brain study. Case Sexa Age (months) Brain Volume (cm3) Case Sexa Age (months) Brain Volume (cm3) B009 M 0.00 518.92 B030 F 2.23 474.45 B011 M 0.00 450.98 B015 M 2.33 702.45 B040 F 0.00 344.38 B020 F 2.60 658.67 B041 F 0.00 331.57 B038 F 3.02 557.40 B045 M 0.00 388.31 B047 M 3.78 662.65 B024 M 0.03 389.26 B023 F 4.80 669.64 B002 M 0.03 341.55 B039 M 4.99 895.41 B003 M 0.03 362.35 B035 F 6.11 1066.78 B010 M 0.03 518.38 B013 M 6.31 814.55 B016 M 0.03 401.68 B014 F 6.70 885.64 B032 M 0.07 318.56 B017 M 7.23 1054.27 B007 M 0.10 365.50 B034 M 7.92 653.30 B027 M 0.13 333.64 B028 M 8.57 934.11 B012 M 0.23 502.82 B006 M 9.77 859.78 B005 M 0.26 403.13 B018 M 10.94 1027.64 B022 M 0.30 258.85 B044 F 11.56 1163.88 B004 M 0.33 359.97 B026 F 12.06 857.82 B019 M 0.62 537.69 B036 M 12.29 978.48 B031 M 0.76 498.18 B042 M 13.63 1143.16 B008 F 0.95 329.72 B037 F 13.63 802.46 B021 F 1.54 507.01 B043 M 14.29 1232.03 B033 F 1.54 427.76 B046 F 14.55 898.34 B025 M 1.77 527.16 B001 M 16.79 1159.42 B029 F 2.17 704.89 a M, male; F, female

PAGE 75

64 Table 4-2. Average brain volume during different period of the first year and a half of life and its characteristics. Age Intervals Mean Brain Volume (cm3) STDa Minb (cm3) Maxc (cm3) Differenced (%) 0 406.83 78.14 331.57 518.92 36 0-3m 451.08 124.35 258.85 704.89 63 3-6m 697.51 141.97 557.40 895.41 38 6-12m 939.99 155.84 653.30 1163.88 44 1-2y 1010.24 167.85 802.46 1232.03 35 a STD, Standard deviation b Minimum brain volume found over the period c Maximum brain volume found over the period d Difference between the maximum and minimum brain volume found over the period relative to the mean

PAGE 76

65 Table 4-3. Sex, age, skull volume for the 29 children used for the skull study. Case Sexa Age (months) Skull Volume (cm3) Case Sexa Age (months) Skull Volume (cm3) S009 M 0.00 101.43 S021 F 1.54 101.24 S011 M 0.00 86.33 S033 F 1.54 136.16 S041 F 0.00 83.68 S025 M 1.77 159.39 S045 M 0.00 78.79 S029 F 2.17 180.10 S002 M 0.03 70.56 S030 F 2.23 131.19 S003 M 0.03 74.53 S038 F 3.02 154.89 S010 M 0.03 101.22 S013 M 6.31 240.24 S032 M 0.07 77.32 S034 M 7.92 259.49 S007 M 0.10 64.23 S028 M 8.57 222.43 S027 M 0.13 90.75 S006 M 9.77 207.42 S005 M 0.26 100.10 S026 F 12.06 241.58 S022 M 0.30 58.17 S042 M 13.63 332.54 S019 M 0.62 116.53 S037 F 13.63 279.85 S031 M 0.76 116.23 S043 M 14.29 344.07 S008 F 0.95 68.46 a M, male; F, female

PAGE 77

66 Table 4-4. Average skull volume during different period of the first year and a half of life and its characteristics. Age Intervals Mean Skull volume (cm3) STDa Minb (cm3) Maxc (cm3) Differenced (%) 0 87.56 9.76 78.79 101.43 22 0-3m 102.89 35.37 58.17 180.10 68 3-6m 154.89 N/A N/A N/A N/A 6-12m 232.60 22.60 207.42 259.49 20 1-2y 299.51 47.68 241.58 344.07 30 a STD, Standard deviation b Minimum skull volume found over the period c Maximum skull volume found over the period d Difference between the maximum and minimum skull volume found over the period relative to the mean

PAGE 78

67 Table 4-5. Sex, age, lung volume for the 20 children used for the lung study. Case Sexa Age (months) Total Lung Volume (cm3) Left Lung Volume (cm3) Right Lung Volume (cm3) Difference between left and right lung (%) LU005 F 0.03 190.52 80.55 109.97 27 LU016 M 0.26 163.12 67.17 95.95 30 LU002 M 1.81 283.04 143.87 139.17 -3 LU003 F 1.97 179.86 69.45 110.41 37 LU011 M 2.96 183.80 80.18 103.62 23 LU017 M 3.78 370.73 169.08 201.65 16 LU020 M 4.60 461.69 209.15 252.54 17 LU004 M 5.36 238.83 113.89 124.94 9 LU006 M 5.72 404.96 179.65 225.31 20 LU001 M 5.88 210.19 102.14 108.05 5 LU008 F 6.31 375.52 185.33 190.19 3 LU012 F 8.64 337.78 138.68 199.1 30 LU010 M 8.71 458.17 227.96 230.21 1 LU009 M 9.23 404.51 182.45 222.06 18 LU013 M 9.33 333.74 156.47 177.27 12 LU014 M 11.17 330.13 162.30 167.83 3 LU007 M 11.79 358.79 154.06 204.73 25 LU018 F 13.60 621.13 297.30 323.83 8 LU015 M 14.95 500.41 237.06 263.35 10 LU019 F 15.67 427.61 188.44 239.17 21 aM, male; F, female

PAGE 79

68 Table 4-6. Average lung volume during different period of the first year and a half of life and its characteristics. A) Total lung volume. B) Left lung volume. C) Right lung volume. A) Age Intervals Mean Total Lung Volume (cm3) STDa Minb (cm3) Maxc (cm3) Differenced (%) 0-3m 200.07 47.47 163.12 283.04 42 3-6m 337.28 108.42 210.19 461.69 54 6-9m 390.49 61.58 337.78 458.17 26 9-12m 356.79 34.27 330.13 404.51 18 1-2y 516.38 97.74 427.61 621.13 31 B) Age Intervals Mean Left Lung volume (cm3) STDa Minb (cm3) Maxc (cm3) Differenced (%) 0-3m 88.24 31.69 67.17 143.87 53 3-6m 154.78 45.34 102.14 209.15 51 6-9m 183.99 44.66 138.68 227.96 39 9-12m 163.82 12.89 154.06 182.45 16 1-2y 240.93 54.53 188.44 297.30 37 C) Age Intervals Mean Right Lung Volume (cm3) STDa Minb (cm3) Maxc (cm3) Differenced (%) 0-3m 111.82 16.37 95.95 139.17 31 3-6m 182.50 63.17 108.05 252.54 57 6-9m 206.50 21.01 190.19 230.21 17 9-12m 192.97 24.92 167.83 222.06 24 1-2y 275.45 43.61 239.17 323.83 26 a STD, Standard deviation b Minimum lung volume found over the period c Maximum lung volume found over the period d Difference between the maximum and minimum lung volume found over the period relative to the mean

PAGE 80

69 Table 4-7. Sex, age, liver volume for the 11 children used for the liver study. Case Sexa Age (months) Liver Volume (cm3) Liver Mass (g) LI005 F 0.03 89.67 93.26 LI007 M 148.8 154.75 LI002 F 2.10 187.35 194.84 LI011 M 4.70 194.51 202.29 LI009 M 5.36 338.91 352.47 LI008 M 5.88 162.11 168.59 LI010 M 11.17 359.75 374.14 LI003 F 11.30 353.67 367.82 LI004 F 11.93 285.55 296.97 LI006 M 12.39 489.33 508.90 LI013 F 15.67 259.72 270.11 0.23 aM, male; F, female

PAGE 81

70 Table 4-8. Sex, age, kidney volume for the 10 children used for the kidney study. Case Sexa Age (months) Total Kidney Volume (cm3) Left Kidney Volume (cm3) Right Kidney Volume (cm3) Difference between Left and Right Kidney (%) Total Kidney Mass (g) K005 F 0.03 21.46 11.67 9.79 -19 22.32 K007 M 0.23 30.65 16.32 14.33 -14 31.88 K001 F 2.17 29.86 14.89 14.97 1 31.05 K012 M 2.46 43.53 22.32 21.21 -5 45.27 K011 M 4.70 39.03 18.43 20.6 11 40.59 K009 M 5.36 36.87 18.74 18.13 -3 38.34 K008 M 5.88 39.55 20.15 19.4 -4 41.13 K003 F 11.30 89.09 46.73 42.36 -10 92.65 K004 F 11.93 50.65 25.61 25.04 -2 52.68 K006 M 12.39 95.52 52.19 43.33 -20 99.34 aM, male; F, female

PAGE 82

71 Table 4-9. Mean brain weight per age period. A) For the CT-based volume data and the Cristy and Ekerman soft tissue density (1.04 g.cm-3). B) For the ICRP Publication 23. C) Difference between the two average masses. A) Age Intervals Mean Brain Weight (g) STDa Number of cases 0 423.11 81.26 5 0-3m 469.12 129.32 22 3-6m 725.41 147.65 4 6-12m 961.11 165.01 8 1-2y 1050.65 174.57 7 B) Age Intervals Mean Brain Weight (g) STDa Number of Cases 0 350.17 4.24 484 0-3m 423.83 16.97 683 3-6m 566.89 46.67 683 6-12m 803.05 106.77 900 1-2y 935.48 54.45 1061 C) Age Intervals Ratio CT/ICRP Difference (%) 0 1.21 20.8 0-3m 1.11 10.7 3-6m 1.28 28.0 6-12m 1.20 19.7 1-2y 1.12 12.3 a STD, standard deviation

PAGE 83

72 Table 4-10. Mean skull weight per age period. This was calculated from the CT-based volume data and the linear interpolation of the Cristy and Ekerman bone tissue densities (1.2 g.cm-3 for a newborn and 1.44 g.cm-3 for 1-year-old and above). Age Intervals Mean Skull Weight (g) STDa Number of cases 0 128.41 35.47 4 0-3m 172.85 59.43 16 3-6m 260.22 N/A 1 6-12m 390.77 37.97 4 1-2y 460.80 38.05 4 aSTD, standard deviation

PAGE 84

73 Table 4-11. Mean lung weight per age period. A) For the CT-based volume data and the Cristy and Ekerman lung tissue density (0.296 g.cm-3). B) For the ICRP Publication 23. C) Difference between the two average masses. A) Age Intervals Mean Lung Weight (g) STDa Number of cases 0-3m 61.95 14.61 5 3-6m 97.36 36.50 5 6-9m 115.59 18.23 3 9-12m 105.61 10.14 4 1-2y 152.85 28.93 3 B) Age Intervals Mean Lung Weight (g) STDa Number of Cases 0-3m 66.17 3.68 93 3-6m 93.70 0.57 105 6-9m 122.52 9.76 127 9-12m 142.06 0.07 112 1-2y 172.89 3.54 162 C) Age Intervals Ratio CT/ICRP Difference (%) 0-3m 0.94 -6 3-6m 1.04 4 6-9m 0.94 -6 9-12m 0.74 -26 1-2y 0.88 -12 aSTD, standard deviation

PAGE 85

74 Table 4-12. Difference between the anthropometric models and the values determined by the growth curves established in this study. A) Newborn stylized model. B) 1-year-old stylized model. C) UF-newborn tomographic model. D) UF-2-month tomographic model. E) 6-month comparison. F) 9-month-tomographic model. A) Organ a Newborn Stylized Model Organ Volume (cm3) Volume determined from the General Equationb (cm3) Differencec (%) C&E Brain 338.00 374.04 10 Bouchet's Brain 371.90 374.04 1 C&E Skull 55.93 87.56 36 Bouchet's Skull 83.50 87.56 5 Left Lung 79.10 67.91 -16 Right Lung 91.90 99.05 7 TLV 171.00 167.42 -2 Average Differenced (%) 2.45 B) Organ a 1-year-old Stylized Model Organ Volume (cm3) Volume determined from the General Equationb (cm3) Differencec (%) C&E Brain 850.00 1005.70 15 Bouchet's Brain 838.40 1005.70 17 C&E Skull 161.80 282.46 43 Bouchet's Skull 252.80 282.46 11 Left Lung 225.00 199.63 -13 Right Lung 259.00 231.12 -12 TLV 484.00 430.85 -12 Average Differenced (%) 13.16 C) Organ a UF-newborn Tomographic Model Organ Volume (cm3) Volume determined from the General Equationb (cm3) Difference (%)c Volume determined from the Specific Equation (cm3)e Difference (%)c Brain 309.86 398.43 22 325.74 5 Skull 97.34 87.79 -11 Left Lung 56.50 78.31 28 Right Lung 66.33 105.43 37 TLV 122.83 183.69 33 Average Differenced (%) 22.08

PAGE 86

75 Table 4-12. Continued. D) Organ a UF-2-month Tomographic Model Organ Volume (cm3) Volume determined from the General Equationb (cm3) Differencec (%) Volume determined from the Specific Equatione (cm3) Differencec (%) Brain 535.97 577.06 7 573.28 7 Skull 166.98 135.62 -23 Left Lung 74.53 111.28 33 Right Lung 80.00 134.12 40 TLV 154.53 245.32 37 Average Differenced (%) 22.42 E) Organ a 6-month Comparison Organ Volume (cm3) Volume determined from the General Equationb (cm3) Differencec (%) Volume determined from the Specific Equatione (cm3) Differencec (%) Brain 535.97 818.25 34 821.79 35 Skull 166.98 211.64 21 Left Lung 74.53 153.92 52 Right Lung 80.00 178.13 55 TLV 154.53 331.84 53 Average Differenced (%) 36.34 F) Organ a UF-9-month Tomographic Model Organ Volume (cm3) Volume determined from the General Equationb (cm3) Difference c (%) Volume determined from the Specific Equatione (cm3) Differencec (%) Brain 948.00 927.80 -2 950.45 0.3 Skull 323.00 251.59 -28 Left Lung 152.00 178.11 15 Right Lung 183.00 205.80 11 TLV 335.00 384.04 13 Average Differenced (%) 14.44 a C&E, Cristy and Ekermans head model (Cristy 1980; Cristy and Eckerman 1987); the lung model is the Cristy and Ekerman s model as well. b the general equation for the brain is the equation (4-3), for the skull the equation (4-6), for the TLV the equation (4-7), for the left lung the equation (4-8), for the right lung the equation (4-9). c relative organ volume differences between the model volume and the growth curve calculation. d model average differences are calculated as indicated in Eq. (4-10). e the specific equation for the brain is (4-4) for C), (4-5) for D), E) and F).

PAGE 87

76 Figure 4-1. Evolution of the brain volume as a function of age over the first year and a half of life.

PAGE 88

77 Figure 4-2. Brain volume comparison between CT and MRI volumetry technique. Evolution of the brain volume as a function of age over the first year and a half of life for the two techniques.

PAGE 89

78 Figure 4-3. Evolution of the brain volume as a function of age over the first year and a half of life for male and female.

PAGE 90

79 Figure 4-4. Evolution of the skull volume as a function of age over the first year and a half of life.

PAGE 91

80 A) B) C) Figure 4-5. 3D reconstruction showing the evolution of the skull volume as a function of age over the first year and a half of life. A) Skull of a newborn with the large fontanels of connective tissues. B) Skull of a 6.31-month-old child, where all the fontanels are still observable, but have been filled by bones except the anterior one. C) Skull of a 13.63-month-old child. The anterior fontanel has almost disappeared. The skull is less circular.

PAGE 92

81 Figure 4-6. Evolution of the total lung volume as a function of age over the first year and a half of life.

PAGE 93

82 Figure 4-7. Evolution of the left and right lung volume as a function of age over the first year and a half of life.

PAGE 94

83 Figure 4-8. Relative difference between left and right lung volume as a function of age over the first year and a half of life. The right lung is compared to the left lung.

PAGE 95

84 050100150200250300350LU001LU002LU003LU004LU005LU006LU007LU008LU009LU010LU011LU012LU013LU014LU015LU016LU017LU018LU019LU020Case #Lung Volume (cm3) Left Lung Right Lung Figure 4-9. Difference between the left and right lung volume for 20 children in their first year and a half of life.

PAGE 96

85 0102030405060K001K003K004K005K006K007K008K009K011K012Case #Kidney Volume (cm3) Left Kidney Right Kidney Figure 4-10. Difference between the left and right kidney volume difference for 10 children in their first year and a half of life.

PAGE 97

86 725.41469.12423.111050.65961.11935.48803.05566.89423.83350.170.00200.00400.00600.00800.001000.001200.0000-3m3-6m6-12m1-2yAge IntervalsBrain Mass (g) Brain CT Brain ICRP 23 Figure 4-11. Brain mass average evolution for CT-based data and for ICRP data per age period over the first year and a half of life.

PAGE 98

87 122.52142.06152.85105.61115.5997.3661.9566.17172.8993.700.0020.0040.0060.0080.00100.00120.00140.00160.00180.00200.000-3m3-6m6-9m9-12m1-2yAge IntervalsLung Mass (g) Lung CT Lung ICRP 23 Figure 4-12. Lung mass average evolution for CT-based data and ICRP data per age period over the first year and a half of life.

PAGE 99

CHAPTER 5 CONCLUSION AND FUTURE WORK A brief summary of the establishment of the organ-specific growth curves will be presented hereafter followed by the conclusions that can be stated from this study. Finally, some suggestions for the future establishment of new organ-specific growth curves will be suggested. Conclusions Over one hundred CT images were successfully segmented for this study and the organ volumes of the brain, the skull, the lungs, the liver, and the kidneys were determined. Some errors, which are essentially due to the image quality and the appreciation of the observer, can exist in this organ volume determination. Image artifacts can be numerous especially with living patients and can cause problems with semiautomatic and automatic segmentation. The presence of tubes or surgery staples inside the patient as well as the equipment surrounding the patient that cannot always be removed for the side of the patient can make the segmentation more difficult. The manual segmentation helped to correct these errors. Organ-specific growth curves were established for the normal development of the brain, the skull, and the lungs. For symmetrical organs the differentiation between the left and right organs (lungs and kidneys) was quantified. Furthermore, gender differentiation was taken into account when sufficient cases were available. The graphs of the organ evolution over time demonstrate a non-linear growth of the organ studied. The great variation that can exist between individuals at the same age was largely 88

PAGE 100

89 observed. Furthermore, the organs do not grow under the same schema. The brain and the skull seem to have a very rapid growth in the first few months of life and a slower one afterwards. The lungs, on the other hand, seem to have a relatively constant growth over the first year of life. However, this curve could have been more accurate with a lot more cases even though the results presented here give a good general trend of organ growth. Furthermore, with more cases the deviation from the average could be quantified over the time. When comparisons are made to reference weight growth curves that were derived primarily from autopsy studies, differences are observed. These differences can be attributed to a combination of factors such as (1) overall increases in individual size and stature over the past several decades, (2) problems with tissue swelling or shrinkage post mortem, and (3) differences in techniques used for organ volume and/or mass determination. The characterization of the existing anthropometric models used for radiation dosimetry purposes was done based on the growth curves of normal organ development established in this study. The results confirmed that the stylized models were close to representing a standard population over a large period of time, but not a specific individual at a specific age. On another hand, the tomographic models were shown to be a lot more specific to the individual it represents and very different from a standard population. Future Works This study should be continued to fulfill a complete characterization of organ volume growth curves. Furthermore, some improvements can be suggested to make this occur. Increasing the number of cases studied would allow the establishment of a

PAGE 101

90 quantitative deviation from the general evolution trend, a better differentiation of all the parameters that can influence the growth of the different organs as well as the growth of the whole body. Relative results would also provide more useful information than absolute results since organ growth can depend on a lot of parameters (mass, ethnicity, gestational time). This might be a very long and time-consuming process since patients who can be considered normal or healthy do not represent the majority of the cases requiring imaging procedures to be performed at the hospital. However, it might be worthwhile not only for anthropometric model construction for radiation dosimetry purpose. The application of cross-sectional image volumetry is very broad. MR imaging of healthy volunteers might be an alternative to consider. Some characteristic organ dimensions and positions should be taken into account to help define model cross-sections. In addition, imaging as many different organs as possible including bones is also one of the quickest future works to achieve. Nonetheless some limitations have to be considered for very small organs, especially if they have a very low image contrast (testes, ovaries). They may be detected, but errors can be large due to their small size and their relative dimensions with the voxel size. Some other organs, like thymus and thyroid, are rarely fully imaged in CT for living patients because most of the time these organs are not of interest and appear in the upper or lower edges of the data set. The thyroid is usually partially imaged at the bottom of a head data set and the thymus at the top of a chest data set. For organs that have a variable size like the stomach, the bladder, the small and large intestine, important differences may be detected from individual to individual and from time to time. For one organ, some categories may apply with regard

PAGE 102

91 to content such as empty, a quarter full, a half full, and full. However, these categories have to be included to a very large pool of values to give accurate results. Furthermore, a better localization of bone marrow within the body over the childhood and the adolescence should be developed since it is highly sensitive to radiation exposures and is located in very high portion within childs body. Data on this specific subject are limited (Cristy 1981; ICRP 1995). A bone sub-segmentation, which would separate the bone marrow from the rest of the bone, can be performed on CT data when the bone reconstruction filter is used. However, for this purpose, if MR images would not be of any help, cadaver studies may be strongly considered. The contrast is not a problem since bone tissue is a high contrast tissue compared to the soft tissues. The changes in bone structure after death are generally lengthy (over a month). Furthermore, long bones like the spine or the arms and legs are seldom fully imaged in most of the procedures on living patients except if they are of particular interest. The arms are positioned above the head of the patient in almost all the examinations to decrease the amount of radiation that would unnecessarily be received by the patient and that would also create scatter. Trunk examinations usually start at the bottom of the neck and stop at the upper thighs or middle tights. The legs are then rarely imaged. Segmentation allows the determination of organ volume. To calculate organ mass, some density statements were made. Only three different densities from Cristy and Ekermans work were used and may not be the most appropriate to use. ICRU Publication 46 (ICRU 1992) contains more organ-specific tissue densities, but not all ages nor both genders are considered as mentioned in Chapter 3. The use of CT number to determine the specific density of each organ for each individual may be a better step

PAGE 103

92 toward the specification of each model to each individual. One limitation might be the contrast agent injected into the patient which obviously changes the measured tissue density. Some phantom experiments should be performed to assess the accuracy of the segmentation method used. Different acquisition modes, segmentation conditions, and observers may influence the final results. The comparison between the gold standard method of fluid displacement done with cadaver organs and CT volumetry would also provide very useful results to improve the technique. Cadaver conditions are different from those of living patients. For most of the existing models, the results acquired from autopsy studies are used as references for living patients, which might not be the best. CT volumetry may be more appropriate for future model construction. Another major point to work on is the improvement of the segmentation method so that it becomes more automatic and less time-consuming. Automatic segmentation would also reduce subjectivity, even though image artifacts that are not uncommon for living patient images would always require validation studies. In a longer-term study, the research can be expanded to healthy as well as non-healthy people to create lifelong organ growth curves, with emphasis on periods of high growth rate: the first year of life and puberty. Puberty should be considered to extend over a large period since it can start and stop at very different ages for different individuals. The disparity is a lot greater than during the first year of life since all the parameters that need to be taken into account during the early life usually accentuate the disparity over time. Another long-term study would be the use of patient characteristics like age, weight, height, head circumference, ethnicity, gestational time, organ evolution,

PAGE 104

93 and health conditions, to develop software that would build models adapted to each individual.

PAGE 105

LIST OF REFERENCES Abbott AH, Netherway DJ, Nieman DB, Clark B, Yamamoto M, Hanieh AC, Moore MH, and David DJ. 2000. CT-determined Intracranial Volume for a Normal Population. Journal of Craniofacial Surgery 11: 211-223. Anguelov Z. 2001. The Decline of the Autopsy. University of Iowa Health Care. http://www.uihealthcare.com/news/currents/vol2issue3/6autopsy.html Updated: June 11th 2001. Bakker J, Olree M, Kaatee R, Lange EED, and Beek FJA. 1998. In Vitro Measurement of Kidney Size: Comparison of Ultrasonography and MRI. Ultrasound in Medicine and Biology 24: 683-688. Blackman KD. 1933. White House Conference on Child Health and Protection called by President Herbert Hoover, Growth and Development of the Child. Part II. Anatomy and physiology. Report of the Committee on Growth and Development. Century. New York and London. Bouchet LG, Bolch WE, Blanco HP, Wessels BW, Siegel JA, Rajon DA, Clairand I, Sgouros G. 2003. MIRD Pamphlet No. 19: Absorbed Fractions and Radionuclide S Values for Six Age-dependent Multi-region Models of the Kidneys. Journal of Nuclear Medicine 44(7): 1113-1147. Bouchet LG, Bolch WE. 1999. Five Pediatric Head and Brain Mathematical Models for Use in Internal Dosimetry. Journal of Nuclear Medicine 40: 1327-1336. Boyd E. 1941. Outline of Physical Growth and Development. Minneapolis: Burgess. Boyd E. 1952. An Introduction to Human Biology and Anatomy for First Year Medical Student. Denver, CO: Child Research Council. Brannan T. 2001. Sigmaplot 2001 for windows. Chicago, IL. SPSS Inc. Brown MS, Feng WC, Hall TR, McNitt-Gray MF, and Churchill BM. 2001. Knowledge-based Segmentation of Pediatric Kidneys in CT for measurement of Parenchymal Volume. Journal of Computer Assisted Tomography 25: 639-648. Brown MS, McNitt-Gray MF, Goldin JG, Greaser LE, Hayward UM, Sayre JW, Arid MK, and Aberle DR. 1999. Automated Measurement of Single and Total Lung Volume from CT. Journal of Computer Assisted Tomography 23: 623-640. 94

PAGE 106

95 Buda FB, Reed JC, and Rabe EF. 1975. Skull Volume in Infants. Methodology, Normal Value, and Application. American Journal of Diseases of Children 129: 1171 1174. Caon M, Bibbo G, and Pattison J. 1999. An EGS4 ready Tomographic Computi onal Model of a 14 year old Female Torso for Calculating Organ Doses from CT Examinations. Phys. Med. Biol. 44: 2213 2225. Chalk River Conference on Permissible Dose. 1949. In: A Conference of Representatives from United Kingdom, Canada, and the United Sta tes. Chalk River, Ontario, Canada. September 29 th and 30 th 1949. RM 10, 1950. Chapman B, O'Callaghan C, and Coxon R. 1990. Estimation of Lung Volume in Infants by Echo Planar Imaging and Total Body Plethysmography. Arch Dis Child 65: 168 170. Clinical Gro wth Charts. 2001. National Center for Health Statistics. Updated: December 30 th 2002. http://www.cdc.gov/growthcharts/ Cristy M. 1980. Mathematical Phantoms Representing Children of Various Ages for Use in Estimates of Internal Dose. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL/NUREG/TM 367. Cristy M. 1981. Active Bone Marrow Distribution as a Function of Age in Humans. Phys. Med. Biol. 26: 389 400. Cristy M and Eckerman KF. 1987. Specific Absorbed Fr actions of Energy at Various Ages from Internal Photons Sources. Oak Ridge, TN: Oak Ridge National Laboratory. Denison DM, Morgan MDL, and Millar AB. 1986. Estimation of Regional Gas and Tissue Volumes of the Lung in Supine Man Using Computed Tomography. Thorax 41: 620 628. Dimbylow PJ. 1996. The Development of Realistic Voxel Phantoms for Electromagnetic Field Dosimetry. P. J. Diwbylow. Chilton, UK: National Radiological Protection Board. Disler DG, Marr DS, and Rosenthal DI. 1994. Accuracy of Volume Mea surements of Computed Tomography and Magnetic Resonance Imaging Phantoms by three dimensional Reconstruction and Preliminary Clinical Application. Investigative Radiology 29: 739 745. Eckerman KF, Cristy M, and Ryman JC. 1996. The ORNL Mathematical Phantom Series. Accessed:August 8 th 2003 http://ordose.ornl.gov/resources/phantom.html Fill U, Zankl M, N P H, and Regulla D. 2002. Adult Female Voxel Models of Different Stature and Photon Conversion Coefficients for Radiation Protection.

PAGE 107

96 Gault D, Brunelle F, Renier D, and Marchac D. 1988. The Calculation of Intracranial Volume Using CT Scans. Childs Nervous System 4: 271 273. Geise RA and O'Dea TJ. 1999. Radiation Dose in Interventional Fluoroscopic Procedures. Applied Radiation and Isotopes 50: 173 184. Gibbs JS, Pujol A, Chen T S, M alcolm AW, and James AE. 1984. Patient Risk from Interproximal Radiology. Oral Surgery, Oral Medecine, Oral Pathology 58: 347 354. Gunduz H, Wu H, Ashtari M, Bogerts B, Crandall D, Robinson DG, Alvir J, Lieberman J, Kane J, and Bilder R. 2002. Basal Gangli a Volumes in First episode Schizophrenia and Healthy Comparison Subjects. Biological Psychiatry 51: 801 808. Hamano K and Iwasaki N. 1990. Volumetric Quantification of Brain Volume in Children Using Sequential CT Scans. Neuroradiology 32: 300 303. Heymsfie ld SB, Fulenwider T, Norlinger B, Barlow R, Sones P, and Kutner M. 1979. Accurate Measurement of Liver, Kidney, and Spleen Volume and Mass by Computerized Axial Tomography. Annals of Internal Medecine 90: 185 187. Hilyer MJC, Snyder WS, and Warner GG. 1972 Estimates of Dose to Infants and Children from a Photon Emitter in the Lungs. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL 4811. Hoe LV, Haven F, Bellon E, Baert AL, Bosmans H, Feron M, Suetens P, and Marchal G. 1997. Factors Influencing the Accura cy of Volume Measurements in Spiral CT: a Phantom Study. Journal of Computer Assisted Tomography 21: 332 8. Horsefield K. 1967. Morphology of the Human Bronchial Tree. Birmingham: University of Birmingham. Hwang JML, Shoup RL, and Poston JW. 1976a. Mathem atical Description of a Newborn Human for Use in Dosimetry Calculations. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL/TM 5453. Hwang JML, Shoup RL, and Poston JW. 1976b. Modification and Additions to the Pediatric and Adult Phantoms. Oak Ridge, TN: O ak Ridge National Laboratory. ORNL/TM 5454. Hwang JML, Shoup RL, Warner GG, and Poston JW. 1976c. Mathematical Description of a One and Five year old Child for Use in Dosimetry Calculation. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL/TM 5293. International Com mission on Radiation Protection 1959. Recommendations of the International Commission on Radiological Protection. Oxford, UK: International Commission on Radiation Protection. Publication 2.

PAGE 108

97 ICRP. 1975. Report on the Task Group on Reference Man. Oxford, UK: International Commission on Radiation Protection. Publication 23. ICRP. 1995. Basic Anatomical and Physiological Data for Use in Radiological Protection: the Skeleton. Oxford, UK: International Commission on Radiation Protection. Publication 70. ICRU. 1992. Photon, Electron, Proton and Neutron Interaction Data for Body Tissues. Bethesda, MD: International Commission on Radiation Units and Measurements. Report 46. IDL Version 5.5, Microsoft Windows (Win32x86). 2001. Boulder, CO. Research Systems, Inc. Jones RM, Poston JW, Hwang JML, Jones TD, and Warner GG. 1976. The Development and Use of a Fifteen-year-old Equivalent Mathematical Phantom for Internal Dose Calculations. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL/TM-5278. Keats TE and Lusted LB. 1978. Atlas of roentgenographic measurement. Chicago: Year Book Medical Publishers. Madirossian G, Tagesson M, Blanco P, Bouchet LG, Stabin M, Yoriyaz H, Baza S, Ljungberg M, Strand S-E, and Bill AB. 1999. A New Rectal Model for Dosimeter Applications. The Journal of Nuclear Medicine 40: 1524-1531. Mahesh M. 2001. Fluoroscopy: Patient Radiation Exposure Issue. Radiagraphics: a Review Publication of the Radiological Society of North America 21: 1033-1045. NCRP. 1989. Exposure of the U.S. Population from Diagnostic Medical Radiation. Bethesda, MD: National Council on Radiation Protection and Measurements. Report 100. NCRP. 1990. Implantation of the Principle of As Low As Reasonably Achievable (ALARA) for Medical and Dental Personnel. Bethesda, MD: National Council on Radiation Protection and Measurements. Report 107. Nipper JC. 2002. Creation of Two Voxelized, Tomographic Models of Pediatric Patients in the First Year of Life. Gainesville: University of Florida. Biomedical Engineering. Nipper JC, Williams JL, and Bolch WE. 2002. Creation of Two Tomographic Voxel Models of Pediatric Patients in the First Year of Life. Phys. Med. Biol. 47: 3143-3164. Noda T, Todani T, Watanabe Y, and Yamamoto S. 1997. Liver Volume in Children Measured by Computed Tomography. Pediatric Radiology 27: 250-252.

PAGE 109

98 Oey I F, Waller DA, Bal S, Singh SJ, Spyt TJ, and Morgan MDL. 2002. Lung Volume Reduction Surgery -A Comparison of the Long Term Outcome of Unilateral vs. Bilateral Approaches. European Journal of Cardio Thoracic Surgery 22: 610 614. Ounsted M, Moar V, and Scott A. 1982. Growth in the First Year of Life: Effects of Sex and Weight for Gestational Age at Birth. Developmental medicine and child neurology 24: 356 365. Pierce RJ, Brown DJ, and Denison DM. 1980. Radiographic, Scintigraphic, and Gas dilution Estimates o f Individual Lung and Lobar Volumes in Man. Thorax 35: 773 780. Poston JW, Snyder WS, and Warner GG. 1975. Age Factor for Dose Rates from Infinite Cloud of a Phantom Emitter. Oak Ridge, TN: Oak Ridge National Laboratory. Rajapakse JC, Giedd JN, DeCarli C, Snell JW, McLaughlin A, Vauss YC, Krain AL, Hamburger S, and Rapoport JL. 1996. A Technique for Single channel MR Brain Tissue Segmentation: Application to a Pediatric Sample. Magnetic Resonance Imaging 14: 1053 1065. Saito K, Wittmann A, Koga S, Ida Y, K amei T, Funabiki J, and Zankl M. 2001. Construction of a Computed Tomographic Phantom for a Japanese Male Adult and Dose Calculation System. Radiat Environ Biophys 40: 69 75. Sgouros S, Goldin JH, Hockley AD, Wake MJ, and Natarajan K. 1999. Intracranial Vo lume Change in Childhood. Journal of Neurosurgery 91: 610 616. Shope TB. 2002. Radiation induced Skin Injuries from Fluoroscopy. FDA / Center of Devices and Radiologicl Health. 08/05/1997. Sinclair D and Dangerfield P. 1978. Human Growth After Birth. Lond on, New York: Oxford University Press. Snyder WS and Cook MJ. 1971. Preliminary Indications of the Age Variation of the Specific Absorbed Fraction for Photons. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL 4720. Snyder WS and Ford MR. 1973. Estimated of Dose Rate to Gonads of Infants and Children from a Photon Emitter in Various Organs of the Body. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL 4903. Snyder WS, Ford MR, Poston JW, and Warner GG. 1976. Estimates of Photon Dose to the Gonads per Micr ocurie day as a Function of the Source Organ and the Age of the Individual. Oak Ridge, TN: Oak Ridge National Report. ORNL 5171. Snyder WS, Ford MR, and Warner GG. 1978. Estimates of Specific Absorbed Fractions for Photon Sources Uniformly Distributed in V arious Organs of a Heterogeneous Phantom. Society of Nuclear Medicine. MIRD Pamphlet No. 5, revised.

PAGE 110

99 Spitzer VM and Whitlock DG. 1989. Atlas of the Visible Human Man. Subbury, MA: Jones and Bartlett. Tsatsanis KD, Rourke BP, Klin A, Volkmar FR, Cicchetti D, and Schultz RT. 2003. Reduced Thalamic Volume in High-functioning Individuals with Autism. Biological Psychiatry 53: 121-129. Urata K, Kawasaki S, Matsunami H, Hashikura Y, Ikegami T, Ishizone S, Momose Y, Komiyama A, and Makuushi M. 1995. Calculation of Child and Adult Standard Liver Volume for Liver Transplantation. Hepatology 21: 1317-1321. Vaidyanathan M, Clarke LP, Heidtman C, Velthuizen RP, and Hall LO. 1997. Normal Brain Volume Measurements Using Multispectral MRI Segmentation. Magnetic Resonance Imaging 15: 87-97. Waitzman AA, Posnick JC, Armstrong DC, and Pron GE. 1992a. Craniofacial Skeletal Measurements Based on Computed Tomography: Part I. Accuracy and Reproducibility. Cleft Palate-Craniofacial Journal 29: 112-117. Waitzman AA, Posnick JC, Armstrong DC, and Pron GE. 1992b. Craniofacial Skeletal Measurements Based on Computed Tomography: Part II. Normal Values and Growth Trends. Cleft Palate-Craniofacial Journal 29: 118-128. Wakabayashi H, Nishiyama Y, Ushiyama T, Maeba T, and Maeta H. 2002. Evaluation of the Effect of Age on Functioning Hepatocyte Mass and Liver Blood Flow Using Liver Scintigraphy in Preoperative Estimations for Surgical Patients: Comparison with CT Volumetry. Journal of Surgical Research 106: 246-253. Warner GG, Poston JW, and Snyder WS. 1974. Absorbed Dose in Male Humanoid Phantoms from External Sources of Photons as a Function of Age. Oak Ridge, TN: Oak Ridge National Laboratory. ORNL-4979. Watanabe Y, Todani T, Noda T, and Yamamoto S. 1997. Standard Splenic Volume in Children and Young Adults Measured from CT Images. Surgery Today 27: 726-728. Williams G, Zankl M, Abmayr W, Veit R, and Drexler G. 1986. The Calculation of Dose from External Photon Exposures Using Reference and Realistic Human Phantoms and Monte Carlo Methods. Phys. Med. Biol. 31: 449-452. Xu XG, Chao TC, and Bozkurt A. 2000. A 2000 VIP-MAN: An Image-based Whole Body Adult Male Model Constructed from Color Photographs of the Visible Human Project for Multi-particle Monte Carlo Calculations. Health Phys. 78: 476-486. Zankl M, Veit R, Williams G, Scheinder K, Fendel H, Petoussi N, and Drexler G. 1988. The Construction of Computer Tomographic Phantoms and their Application in Radiology and Radiation Protection. Radiat. Environ. Biophys. 27: 153-164.

PAGE 111

100 Zankl M and Wittmann A. 2001. The Adult Male Voxel Model "Golem" Segmented from Whole-body CT Patient Data. Radiat Environ Biophys. 40: 153-62. Zubal IG, Harell CR, Smith EO, and Smith AL. 1996. Two Dedicated Sotfware, Voxel-based, Anthropomorphic (torso and head) Phantoms. P. J. Dimbbylow. Chilton, UK: National Radiological Protection Board.

PAGE 112

BIOGRAPHICAL SKETCH Nelly Aline Volland was born in Bourgoin-Jallieu, France, in August of 1979. She graduated high school from Lyce Jean Moulin in Albertville, France, in June 1997. She then joined the pre-engineering school Cycle Prparatoire Polytechnique in Nancy (CPP, France) for two years, from which she obtained her entrance to engineering school. She achieved her engineering degree in physics at the Ecole Nationale Suprieure de Physique de Grenoble (ENSPG, France) in September 2002. 101


Permanent Link: http://ufdc.ufl.edu/UFE0001375/00001

Material Information

Title: Organ Volumes in Pediatric Patients Assessed via Computed Tomography Image Segmentation
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0001375:00001

Permanent Link: http://ufdc.ufl.edu/UFE0001375/00001

Material Information

Title: Organ Volumes in Pediatric Patients Assessed via Computed Tomography Image Segmentation
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0001375:00001


This item has the following downloads:


Full Text












ORGAN VOLUMES IN PEDIATRIC PATIENTS
ASSESSED VIA COMPUTER TOMOGRAPHY
IMAGE SEGMENTATION















By

NELLY A. VOLLAND


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2003

































Copyright 2003

by

Nelly A. Volland































To Sophie,
Dead on June 5th 1996
from leukemia at the age of 17.
She will never get the chance to see
what a College student life looks like.















ACKNOWLEDGMENTS

The author would like to thank all the people who have made this project possible.

The first thanks go to Dr. Wesley Bolch for the opportunity, the finances, and the

computing resources he offered to develop this project.

The author would also like to greatly thank Dr. Williams for the time he spent to

help find the useful anatomy data needed for the project, check the accuracy of the

results, and answer all the anatomy questions. She is also thankful to Dr. Arreola for the

time he spent to explain the functioning of computer tomography imaging technique.

Special thanks also go to the CT technologists at Shands Teaching Hospital for the

kindness, the understanding, and the help they gave during the time spent looking for data

in the CT control room. Special thanks also go to Joshua Nipper for the segmentation

program he developed, which was very useful during this project and Choonik Lee for all

his explanation he could give about this program.

Finally, the author would like to thank her family and friends for their support and

guidance during this work, which would not have been possible without all of them.
















TABLE OF CONTENTS
page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .................... ................. .............. .. ... ........ vii

LIST OF FIGURES ..................................... ................ .......... viii

A B ST R A C T ................. ................................................................................... ........

CHAPTER

1 IN TR O D U C TIO N ......................................................................... .... .. ........

2 B A C K G R O U N D .................... .... ................................ ...... ........ ...............

Stylized Models .................... .......................... ...........
Tom graphic M odels....................................................9
P pediatric D ata Sources ......................................................................... ......... ........ 12

3 M ATERIALS AND M ETHOD S ........................................... .......................... 23

G general Selection and Segm entation................................... ....................................23
CT Data Selection.................. ............................23
Segm entation T technique ........................................ ................. ............... 25
Organ-specific Selection and Segmentation............... ...................................28
B rain ............................................................................. 2 8
S k u ll ...................................................................................................... 3 0
L u n g s ............................................................................. 3 2
L iver and K idneys .............................. .................... .. .............. .............34
G rap h ical A n aly sis........... .............................................. .................. .......... .. ..... .. 3 6

4 RESULTS AND DISCU SSION ............................................ ........................... 46

O rgan G row th Curves............................................................ .. ............... 46
B rain ............................................................................. 4 7
S k u ll ................................................................................................................ 4 9
L ung s .........................................................................................................5 1
L iv e r ................................................................5 3
K idneys.............................................................54
Comparisons with ICRP Reference Values ............................... ......................54


v









Characterization of Stylized and Tomographic Models .........................................56
Stylized M models .................................... .......................... .... ....... .57
T om graphic M odels............................................................... ............... 58

5 CONCLUSION AND FUTURE W ORK ........................................ .....................88

C o n c lu sio n s.................................................... .................. 8 8
Future W works ........................................... ........................... 89

L IST O F R EFE R E N C E S ............................................................................ .............. 94

BIOGRAPHICAL SKETCH ............................................................. ............... 101
















LIST OF TABLES


Table pge

1-1 Dose equivalent threshold above which different deterministic effects can occur
for different organs when irradiated with x-rays ......................................................4

2-1 Existing whole-body voxelized models from child to adult ..................................16

3-1 Head CT exam characteristics for brain and skull studies ....................................38

3-2 CT exam characteristics for lung study ........................................ ............... 40

3-3 CT exam characteristics for liver and kidney studies............... ...............41

4-1 Sex, age, brain volume for the 47 children used for the brain study........................63

4-2 Average brain volume during different period of the first year and a half of life
and its characteristics .................................... .............. ...... ............ 64

4-3 Sex, age, skull volume for the 29 children used for the skull study......................65

4-4 Average skull volume during different period of the first year and a half of life
and its characteristics .................................... .............. ...... ............ 66

4-5 Sex, age, lung volume for the 20 children used for the lung study........................67

4-6 Average lung volume during different period of the first year and a half of life
and its characteristics ............................... ..... .............. ...... ............ 68

4-7 Sex, age, liver volume for the 11 children used for the liver study .......................69

4-8 Sex, age, kidney volume for the 10 children used for the kidney study ..................70

4-9 M ean brain w eight per age period....................................... ......................... 71

4-10 M ean skull w eight per age period ........................................ ........................ 72

4-11 Mean lung weight per age period.......................... ........... ................. 73

4-12 Difference between the anthropometric models and the values determined by
the growth curves established in this study ................................... ............... 74
















LIST OF FIGURES


Figure page

1-1 Photos of necrosis ..... ........ ........................... ......... ....... .......... .. ..

2-1 Internal view of the stylized model adapted from the MIRD Pamphlet No. 5. .......17

2-2 External view and cross-sectional view of the Cristy and Ekerman's stylized
m o d el series s .................................................... ................ 1 8

2-3 Head model for the six stylized models as developed by Bouchet et al. ................19

2-4 Posterior view of the rectal model for the six stylized model as developed by
M arirossian et al....................... ........................... .... ........ ......... 20

2-5 Models used in radiation dosimetry to assess dose profile received by pediatric
patient undergoing diagnostic or interventional examinations .............................21

2-6 Height growth rate curve from birth to 18 years of age adapted from the growth
in h e ig h t ....................................................................... 2 2

3-1 Patient distribution of the brain study by age and sex ............. ............................42

3-2 Patient distribution of the skull study by age and sex................... ....... ........43

3-3 Patient distribution of the lung study by age and sex ............ ............. .............44

3-4 Patient distribution of the liver and kidney studies by age and sex ....................45

4-1 Evolution of the brain volume as a function of age over the first year and a half
of life. ............. ............. .......... .... .......... .... .. ..................................... 76

4-2 Brain volume comparison between CT and MRI volumetry technique ..................77

4-3 Evolution of the brain volume as a function of age over the first year and a half
of life for male and female. ........................... ................. ............... 78

4-4 Evolution of the skull volume as a function of age over the first year and a half
of life. ...................... ............... .......... ............... 79

4-5 3D reconstruction showing the evolution of the skull volume as a function of
age over the first year and a half of life ............ .............................. ............... 80









4-6 Evolution of the total lung volume as a function of age over the first year and a
h alf of life. .............. ... ......... ................................................ ............ ....8 1

4-7 Evolution of the left and right lung volume as a function of age over the first
y ear and a half of life ......... ................................................................. 82

4-8 Relative difference between left and right lung volume as a function of age over
the first year and a half of life ..... .................... ....... .............. 83

4-9 Difference between the left and right lung volume for 20 children in their first
year and a half of life ..... ...... ...................... ...... ........................... ... 84

4-10 Difference between the left and right kidney volume difference for 10 children
in their first year and a half of life. ................................. ............... 85

4-11 Brain mass average evolution for CT-based data and for ICRP data per age
period over the first year and a half of life. ....................................................... 86

4-12 Lung mass average evolution for CT-based data and ICRP data per age period
over the first year and a half of life. ....... ............................... ................. 87















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of
the Requirements for the Degree of Master of Science

ORGAN VOLUMES IN PEDIATRIC PATIENTS ASSESSED VIA COMPUTER
TOMOGRAPHY IMAGE SEGMENTATION

By

Nelly A. Volland

December, 2003

Chair: Wesley E. Bolch
Major Department: Biomedical Engineering

Accurate assessments of the radiation absorbed dose to internal organs are needed

to assess some level of risk to pediatric patients undergoing radiological procedures.

Very young children are considered to be more sensitive to radiation exposures than

adults. They can also be uncooperative during the examinations resulting in longer

exposures. Furthermore, their smaller size exposes more of their anatomy to the x-ray

beam with corresponding increases in their effective dose. Their risk of developing

cancer after radiation exposure is also considered to be greater since pediatric patients

have a longer post-exposure life expectancy than adult patients. As a consequence, the

improvement of quantitative methods for determining organ doses received by pediatric

patients undergoing radiological procedures requires the development of anthropometric

models that better represent pediatric anatomy. However, very few anthropometric

models for children under one year of age have been built to conduct accurate radiation

dosimetry calculations. Furthermore, few references are available on organ parameters









necessary to properly characterize the few existing models and to develop new models

even though organ development is relatively highest during the first year of life.

In the present study, a segmentation technique using cross-sectional computed

tomography (CT) images was used to establishment organ-specific growth curves for

children in their first year and a half of life. In-vivo organ volumes of the brain, the skull,

the lungs, the liver and the kidneys were determined based on the number of voxels

selected to define the organ and the image voxel characteristics. The organ volume was

plotted against the age of the child to determine general organ evolution trends and to

provide the needed information to better characterize existing anthropometric models and

to help the development of new models in the first year of life. Organ masses were also

calculated using reference tissue densities taken from anthropometric model descriptions

from the Oak Ridge National Laboratory (ORNL). Some comparisons with the ICRP

Publication 23 reference values were made since this report has been used to build

anthropometric models for radiation dosimetry.

Finally, the characterization of the existing anthropometric models was performed

using the growth curves established in this study. The results confirmed that the stylized

models were close to a standard population over a large period of time, but not a special

individual at a specific age. On the contrary, the tomographic models were shown to be a

lot more specific to the individuals they represent and very different from a standard

population. The use of cadaver versus living patient to build anthropometric models was

also discussed.














CHAPTER 1
INTRODUCTION

Medical x-rays represent the main man-made source of radiation and the greatest

contribution to radiation doses received by the general public (NCRP 1989). At

sufficiently high radiation doses, patients undergoing interventional radiological

examinations may be subjected to two different types of effects: deterministic effects and

stochastic effects.

Deterministic effects mainly occur within the skin, the first and largest organ of the

body that the radiation beam will irradiate as shown on the Fig. 1-1. They can also affect

eyes or bones during neuroradiology surgery and lungs during cardioradiology surgery

because these organs can remain in the primary beam field during the entire intervention.

These effects are deterministic because there exists a threshold dose, generally given by

the entrance skin dose (ESD), above which they occur and also because they can be

quickly detected, even though they may not appear instantaneously. Table 1-1 gives

some indications about deterministic effects depending on the dose received by the

patient (Geise and O'Dea 1999; Mahesh 2001).

For stochastic effects, a threshold dose may not exist since these effects are only

related to the probability of developing cancer carcinogenesiss effects) or giving deficient

genes to the descendant (hereditary effects). The higher the dose, the more probable the

effects are. However, even though the dose received during a single examination

represents a low risk, the exposure is cumulative. Determining the total radiation dose









for patients undergoing multiple examinations becomes important for further calculations

of radiation risk assessment.

For very sick patients for whom the diagnosis from imaging examinations can

make the difference between life and death, the number of examinations can be

numerous, yet necessary. These examinations can be analyzed to minimize the risk by

suggesting changes in imaging technique factors to keep the dose received by the patient

As Low As Reasonably Achievable (ALARA principle) without losing image quality

(NCRP 1990).

The Pediatric Organ Dose (POD) project developed at UF is oriented toward the

improvement of quantitative methods for determining organ doses received by pediatric

patients undergoing radiological procedures. Pediatric patients are especially at risk

because they are ten times more sensitive to radiation than adults. They can be

uncooperative during the examinations since they do not always understand what is

happening. This may force the technologists and radiologists to prolong the procedure, to

repeat the procedure, or even both. The generally smaller size of pediatric patients also

exposes more of their anatomy to the x-ray beam. This also increases the effective dose

received. Since they have a longer post-exposure life expectancy than adults, their risk of

developing cancer after radiation exposure is elevated. For all these reasons, trying to

understand x-ray examination and to measure radiation dose profile as accurately as

possible to assess some level of risk for x-ray imaging for pediatric patients is one of the

POD project tasks. To achieve this goal, the development of anthropometric models that

better correspond to patients' anatomical characteristics is of prime importance.









The development of both physical and computed anthropometric models started

when people wanted to quantify radiation hazards and determine necessary protection

methods. The stylized models were the first models developed based on simple 3D

analytical geometry and on available references of the human body. These models have

been used for years for radiation dosimetry purposes. However, tomographic models

have become available with the apparition and the improvement of new imaging

technologies. These latter models have generated a need for more specific information

about the human body to better characterize them. Very few detailed data and references

have been published on organ volumes in pediatric patients against which tomographic

model data can be compared to, especially absolute organ volumes and masses at a

specific age. These comparisons become crucial during the first year of life, since this

period corresponds to one of the highest growth rates of the human developmental period.

Establishing organ-specific growth curves over this period would allow a better

characterization and utilization of the existing stylized and tomographic models, as well

as a new database for construction of new models at a given pediatric age. The

segmentation of computed tomography (CT) data set can provide the needed information

to generate the desired growth curves. The study described hereafter provides the

establishment of growth curves over the first twelve months of life for the brain, the

skull, the lungs, the liver, and the kidneys, as well as some applications for the existing

models.









Table 1-1. Dose equivalent threshold above which different deterministic effects can
occur for different organs when irradiated with x-rays. These thresholds are
average doses. They can differ more or less from one individual to another. In
comparison, the population in the United States of America is exposed on
average to the natural background radiation at a dose equivalent rate of
approximately 3 mSv/year.
Single-Dose
Threshold Effect
(Sv)
2 Deterministic effects can be detected
Temporary erythema
Temporary hair loss
Cataract
6 Basal skin cells partially damaged
Long term erythema
Temporary to permanent hair loss
Vision impairment
7 Pneumonitis for large lung area exposed
10 Dry desquamation
18 Basal skin cells completely destructed
30 Possible bone changes in young children
(Dose per unit exposure in bone is about four times as
high as for soft tissue in the range of X-ray used)






5










A) B)

Figure 1-1. Photos of necrosis. A) Skin injury evidencing skin necrosis about 18 to 21
months following fluoroscopy procedures. B) Skin injury about 16 to 21
weeks after the fluoroscopy procedures. A small-ulcerated area is present
(Shope 2002).














CHAPTER 2
BACKGROUND

Accurate dose profiles in human anatomy and methods of simulation of radiation

transport throughout the body require relatively high definition human body

representation. Anatomical modeling for radiation dosimetry was formalized for the first

time in 1949 at the Chalk River Conference on Permissible Dose (Chalk River

Conference on Permissible Dose 1949). Since this first publication, physical as well as

computed models, named as stylized models, have been developed. However, with the

development of new computer techniques, tomographic or voxelized models have also

started to become available since the late 1980s. This evolution as well as the need for

more and more detailed studies has generated a need for more specific information about

the human body than the one found in the actual published data and references.

Stylized Models

The first Standard Man data presented in 1949 (Chalk River Conference on

Permissible Dose 1949) described a standardized set of biological parameters that could

be used to calculate permissible levels of exposure to workers (adult male). In 1959 after

a decade of modifications and improvements the Standard Man was published in the

International Commission on Radiation Protections (ICRP) first report (ICRP 1959).

Updated in 1975 in the ICRP Publication 23 (ICRP 1975), the Standard Man became the

Reference Man (adult male). Biological as well as physiological data were included in

this report according to published data primarily derived from autopsy work. Elemental









composition for various tissues was also summed up. Some data are available for women

and children like masses and organ size, but there are very few.

The Medical Internal Radiation Dose (MIRD) Committee of the Society of Nuclear

Medicine published its Pamphlet No.5 revised in 1978 (Snyder et al. 1978). In this

report, Snyder and his co-workers introduced their latest development done on

anthropometric models for internal radiation dosimetry mainly based on the ICRP

Publication 23. Simple 3D geometry like ellipsoids, truncated cones, and spheres that

better mimic the organ size and shape were used to define this model as shown on Fig. 2-

1. Only the male adult was presented.

The design of the first pediatric phantoms for 0, 1, 5, 10, and 15 years old started

during the same period. Unfortunately, reference data on children were not readily

available. As a consequence, the first pediatric models were based on a similitude rule

(Hilyer et al. 1972; Poston et al. 1975; Snyder and Cook 1971; Snyder and Ford 1973;

Snyder et al. 1976; Warner et al. 1974). This rule transformed the adult phantom into

pediatric phantoms using scaling factors in the three dimensions of the space to

approximate the dimensions of a child at a given age. This technique produced phantoms

that were not realistic pediatric patients. All organs do not have the same evolution

throughout the developmental period. The large head, the shorter legs, the thicker trunk,

and the large thymus and adrenals of children were completely dismissed in these earlier

models. An improvement over these so-called 'similitude phantoms' was done soon after

to better represent pediatric patients (Hwang et al. 1976a; Hwang et al. 1976b; Hwang et

al. 1976c; Jones et al. 1976). Even though the basic type equations to describe the organs

were kept, the organ size, shape, and position were defined according to all the









anatomical references available at the time. In the early 1980's, Cristy and co-workers

created a different set of models for dosimetry purposes (Cristy 1980; Cristy and

Eckerman 1987). While dimensions of the trunk, the head, and the legs were determined

from anthropological data and the organ volumes from ICRP Publication 23 (1975), the

similitude rule was used for the organ shape and localization because of a lack of detailed

data. Fig 2-2 shows this set of models.

When the earliest models have been abandoned, the Cristy and Eckerman's models

(adult male, 15-year-old child that also represents the adult female, 10-, 5-, 1-year-old

children, and newborn) became the unofficial mathematical models for radiation

dosimetry. They were upgraded again by Cristy and Eckerman in 1996 (Eckerman et al.

1996). Further improvements have also been made still based on the ICRP Publication

23 recommendations and depending on specific needs. A new head model for the six

mathematical models was adopted in 1999 (Bouchet et al. 1999). This new version of the

head gives more realistic design of the head and of the neck. Bouchet et al. developed

this modified set of heads in response to a demand of more realistic dose calculations for

radiopharmaceuticals used in neuroimaging as shown on Fig 2-3. A new rectal model

has also been proposed in 1999 (Madirossian et al. 1999) for all the six mathematical

models. This revised version of the lower part of the colon includes the rectum, the

urinary bladder, and the prostate as shown on Fig 2-4. Dose calculations for

radiopharmaceuticals used in brachytherapy and external beam irradiation needed this

modified set of the lower part of the colon to be more accurate. A new kidney model has

been adopted in 2001 (Bouchet et al. 2003). This revised version of the kidneys

describes the specific substructures of this major excretory organ without modifying the









outer dimensions of the existing model. The demand for more detailed local regional

doses to specific kidney substructures for beta-particle and low-energy photons

irradiation have become of greater and greater importance than only single-region

average kidney doses and required the development of this modified set of the kidneys.

The stylized models have been developed to represent 'reference' patients

following the ICRP publication 23 recommendations and using geometry that

approximates as closely as possible the average human body. Small shape variations

from these standards are presumed to have small effects on dose calculations and would

have necessitated more complex computer programs. Therefore, they were not taken into

account to build the models. As a consequence, these models do not represent the patient

with a high degree of accuracy. They are hermaphrodite even though a gender

specification can still be made depending on the purpose of the study. Finally, they do

not represent every age.

With the improvement of computational techniques associated to medical imaging,

and the appearance on the market of powerful and user friendly image analysis software,

some 'new generation' models have been introduced. These new models are called

tomographic or voxelized models.

Tomographic Models

Tomographic or voxelized models can give more realistic representation of human

body as they are developed based on the segmentation of medical images of real patients.

The complicated three-dimensional shape of patient organs can be better defined in these

models than they are in stylized models. This means that the tomographic models would

be more adapted than stylized models to better fit patient characteristics at various ages,









especially for pediatric patients who have rapid morphologic changes in a short period of

time at different rates and starting at different stages.

Several different research groups have already developed some high-resolution

tomographic models mainly for adult patients, but also for some pediatric patients, as

presented on Table 2-1. They create these models of labeled voxels using three-

dimensional data essentially from computed tomography (CT) or magnetic resonance

imaging (MRI).

In each image data set, each voxel composing the data set is assigned the average

characteristic (x-ray attenuation for CT or nuclear magnetic resonance signal for MRI) of

the specific region of interest it represents. The smaller the voxel, the more

homogeneous its density is and the more accurate the model is, particularly for small

organs like ovaries, testes, and adrenals. However, these data do not allow their direct

use. The gray level values on the images for different regions of interest overlap each

other. Using segmentation techniques allow each voxel to be labeled as belonging to a

specific organ, to describe a model, and to define input files suitable for dosimetry

transport codes. However, spatial and contrast resolution are most of the time not high

enough to perform an entirely automatic segmentation.

MR appears to offer better contrast for soft tissues, but the bone tissue definition

needed for radiation dosimetry purpose does not appear as good as it does with CT. In

addition, MR requires a longer acquisition time and a different reconstruction technique

than CT that makes it more sensitive to patient motion artifacts and can severely limit its

application to pediatric patients.









The voxelized model was introduced for the first time in the middle of the 1980's

(Gibbs et al. 1984; Williams et al. 1986). The two first models created from CT data

were published by Zankl et al. in 1988. These models, respectively called BABY and

CHILD, were created from a whole-body scan of an 8-week-old female cadaver and a 7-

year-old female child who was to undergo a whole-body radiation therapy for leukemia

(Zankl et al. 1988). Later, the same team created a tomographic model of a 38-year-old

man, called GOLEM, to approximate the "Reference Man" of the ICRP publication 23

(Zankl and Wittmann 2001). In 1996, Dimbylow created the whole-body voxelized

phantom of an adult male adult based on MRI data (Dimbylow 1996). Xu et al. (2000)

published the most detailed tomographic voxel model. It represents a male adult referred

to as the VIP-man (Xu et al. 2000). It was created from high-resolution photographic

images of the Visible Man, published in 1989 by Spitzer and Whitlock (Spitzer and

Whitlock 1989). However, the model has been said to be too large to represent the

average population. More recently OTOKO (Saito et al. 2001), IRENE and DONNA

(Fill et al. 2002) were created from CT data. OTOKO represents an adult Japanese man

and IRENE and DONNA two adult females. Two baby voxel phantoms of a 6-day-old

female and a 2-month-old male were developed at the University of Florida based on CT

data (Nipper et al. 2002). Fig. 2-5 presents the UF-newborn tomographic model with the

stylized newborn model. Partial body phantoms have also been developed like the head-

torso phantom created in 1994 from CT data and the high-resolution head phantom from

MRI images of a healthy volunteer in 1996 (Zubal et al. 1996). ADELAIDE, a 14-year-

old trunk model was created (Caon et al. 1999).









If one advantage of the tomographic models is that they are more realistic that

stylized models, one disadvantage is that they are more specific. A tomographic model

represents the particular characteristics of a specific patient, but not a standard or an

average person. As opposed to the stylized model, the tomographic model is built first

and then compared to references and also to stylized models. These comparisons help to

identify the characteristics of the models in regards to the average population at the same

age. However, the characterization of the tomographic models already existing and the

construction of other models at various ages in this period can be difficult, because few

detailed references are available on organ growth trend especially for young children.

Furthermore, over the first year of life, the growth rate is one of the highest of the entire

developmental period, as shown on Fig 2-6 for the height (Sinclair and Dangerfield

1978). This growth is not linear.

Pediatric Data Sources

Whereas patient's height, weight, or head circumference can be measured easily

and directly, organ volume or organ mass determination is more difficult. Autopsies and

fluid displacement studies have been the "gold standard" methods to establish human

body references for many years. Autopsy rates were still above 50% in the 1940s in the

US because it was almost the only technique to study anatomy, histology, and disease

characterization. However, by 1985 this rate declined to 14% in teaching hospitals and

even to less than 9% in non-teaching hospitals (Anguelov 2001). The reasons of this

decline are numerous. To perform an autopsy today authorizations from the families are

required and most of the time they are difficult to obtain. These studies are also

expensive to perform and the infant mortality rate for children in their first year of life

has decreased during the last decades. Furthermore, other imaging techniques have









become available that are less evasive to the deceased. X-ray imaging and then

ultrasound imaging have been used as the accuracy of autopsy investigations has

questioned due to post-mortem tissue changes. Lung tissue changes relatively quickly

after death. A progressive but fast fluid accumulation in the lungs postmortem can

influence the measurements as mentioned in the ICRP Publication 23. The consequences

are a lung volume decrease because of air loss and a lung mass increase because of fluid

accumulation postmortem. The liver also contains a high amount of blood whose

modification after death can alter its volume determination. Nonetheless, the accuracy of

these 2D modality techniques (e.g., radiographs and ultrasound) can also be underlined.

The need of simple volume approximations to go from a 2D to a 3D image is obvious.

Often organ volumes are overestimated or underestimated (Bakker et al. 1998).

However, radiologists still read 2D images for diagnosis and still use published organ

measurement tables that give organ characteristic dimensions such as lengths, widths, and

depths (Keats and Lusted 1978) measured on these 2D images. Volume measurements,

when needed, are usually taken from these 2D dimensions.

For these reasons, the stylized models have been built according to the references

established with the autopsies and 2D-to-3D calculations, approximations, and

interpolations. As mentioned previously, the references used to build these models are

indicated in ICRP Publication 23. In this report, the data were taken from different

sources. For example, the brain and kidney data were essentially taken from a report of

the committee on growth and development of 1933 (Blackman 1933). The data are

indicated to be acquired from autopsy work and 2D x-ray imaging. The original source is

nevertheless not specified. The outline of physical growth and development published in









1941 (Boyd 1941) as well as from the Introduction to Human Biology and Anatomy for

first year Medical Student published in 1952 (Boyd 1952) were used for the lung and

liver data. These books summarized Scammon's team autopsy work performed in the

late 1920s.

The improvement of computer technology and the development of cross-sectional

and 3D imaging techniques have brought new methods for building anthropometric

models that are also useful for organ volume and mass determination. These techniques

are able to include the two fundamental conditions (living patients and 3D data) that

make measurements of organ volumes and masses closer to reality than prior techniques.

Their accuracy and precision have been largely tested and underlined (Disler et al. 1994;

Hoe et al. 1997). They started to be used for organ volume and mass determinations in

the late 1970s and have proven to be useful for patient diagnosis and therapy (Gault et al.

1988; Hamano and Iwasaki 1990; Heymsfield et al. 1979). However, they have not been

used in a systematic manner since high technology skills are required to develop

programs that allow quick and easy organ volume and mass determinations.

A lot of research has been oriented toward the development of more and more

automatic and powerful image segmentation programs. Nonetheless, the different body

characteristics and the variety of images techniques usually make studies very specific

(Brown et al. 2001; Brown et al. 1999; Rajapakse et al. 1996; Vaidyanathan et al. 1997).

Aging and diseases are more often the focus of these studies than young and healthy

children (Chapman et al. 1990; Gunduz et al. 2002; Oey et al. 2002; Tsatsanis et al.

2003; Wakabayashi et al. 2002). Furthermore, the entire organ volume is rarely

calculated. The recruitment of people who undergo imaging with no underlying disease









is also a complex time, ethical, financial, and practical issue. However, craniosynostosis,

which is the early fusion of skull bones in the infancy resulting in an abnormal skull and

brain growth, is not uncommon for children. Intracranial volume calculation using 3D

imaging has become of more and more interest to study the abnormality. 3D imaging can

help to better state intracranial volume evolution, skull surgery, and post-surgery follow-

up of the disease. The comparison of the data acquired with a normal growth curve can

then be done. Unfortunately, normal growth curves are rare (Abbott et al. 2000; Sgouros

et al. 1999). Standard liver volume calculation using CT scans has also appeared to be an

accurate and reproducible method that can be beneficial in the assessment of several liver

diseases and liver transplantations (Noda et al. 1997; Urata et al. 1995). Standard spleen

volume calculations have also been developed due to the increasing number of

hypersplenism in children (Watanabe et al. 1997).

Overall, few studies have previously performed on whole organ volume

calculations using 3D imaging for children in their first year of life, particularly healthy

children. Establishing standard organ growth curves for children under one year of age

using CT scans segmentation will provide the needed information to better characterize

models used in radiation dosimetry.






16


Table 2-1. Existing whole-body voxelized models from child to adult.
In-plane Slice
# of
Model Gender Age Resolution Thickness slices Modality Author
(mm) (mm)
UF-Newborn F 6-dy 0.59x0.59 1 485 CT Nipper et al.
UF-2-mth M 2-mth 0.49x0.49 1.25 438 CT Nipper et al.
BABY F 2-mth 0.85x0.85 4 142 CT Zankl et al.
CHILD F 7-yo 1.54x1.54 8 144 CT Zankl et al.
IRENE F 32-yo 1.87X1.87 5 -326 CT Fill et al.
VIP-MAN M 38-yo 0.33x0.33 1 1878 Color Photos Xu etal.
GOLEM M 38-yo 2.08x2.08 8 220 CT Zankl et al.
DONNA F 40-yo 1.87X1.87 10 -176 CT Fill et al.
OTOKO M unknown 0.98x0.98 10 -170 CT Saito et al.
NORMAN M unknown 1.88x1.88 3 -587 MRI Dimbylow









































UPPER LARGE
INTESTINE





URINARY BLADDER


INTESTINE

LARGE INTESTINE


Figure 2-1. Internal view of the stylized model adapted from the MIRD Pamphlet No. 5.











LEFT


RG4HT


Newborn, 1-, 5-, 10-, 15-year-old (adult female), and adult male


Figure 2-2. External view and cross-sectional view of the Cristy and Ekerman's stylized
model series. Adapted from (Cristy 1980; Cristy and Eckerman 1987).






19












A) B)














C) D)














E) F)



Figure 2-3. Head model for the six stylized models as developed by Bouchet et al. A) The
newborn. B) The 1-year-old. C) The 5-year-old. D) The 10-year-old. E) The
15-year-old that also represent the adult female. F) The adult male model.












SI, Small Intestine
DC, Descending Colon
c S, Sigmoid Colon
R, Rectum
UB, Urinary Bladder
P, Prostate












10
UB











Figure 2-4. Posterior view of the rectal model tor the six stylized model as developed by
Marirossian et al. Adapted from (Madirossian et al. 1999)











A)
























Figure 2-5. Models used in radiation dosimetry to assess dose profile received by
pediatric patient undergoing diagnostic or interventional examinations. A)
newborn stylized model. B) 6-day-old tomographic model (Nipper et al.
2002).









22

20-

" 18




E

10
.c 1-

_I


6

4
1 11 J p i I I I I L I
B 1 2 4 6 8 10 12 14 16 18
Age (years)
Figure 2-6. Height growth rate curve from birth to 18 years of age adapted from the
growth in height (Sinclair and Dangerfield 1978).














CHAPTER 3
MATERIALS AND METHODS

To establish organ-specific standard growth curves for the brain, the skull, the

lungs, the liver, and the kidneys, the first task was to collect the in-vivo cross-sectional

CT images of children under one year-of-age and to select the more appropriate ones

according to the criteria of this study (e.g., normal pediatric anatomy). The second task

was the segmentation of the organs to determine their volume and mass in order to build

the growth curves.

General Selection and Segmentation

CT Data Selection

The general CT data selection could be divided in three successive steps: (1) the

selection based on patient age, (2) the selection based on the patient's health condition,

and (3) the selection based on imaging sequences themselves.

In the first step, all data sets chosen for this study came from CT exams of living

patients in their first year and a half of life. The age of the patient was calculated from

the date of birth to the examination day. The exams were performed at Shands Teaching

Hospital in Gainesville, Florida, between September, 2002 and June, 2003. The data

were all acquired on one of four different scanners. The two first scanners were two

identical GE light speed 4-slice scanners with a scan field of view (SFOV) of 250 x 250

mm2. Then, two identical Siemens-Somaton sensation 16 scanners replaced them with a

SFOV of 500 x 500 mm2. Each image was encoded in a 512 x 512 pixel matrix for both

scanners. The protocols used to perform the exams were the ones required for the









specific purposes of the study ordered by the physician. As the study was retrospective in

nature, no decisions on the imaging techniques were made such as obtaining the thinnest

slice thickness achievable for a better axial resolution. Cadavers would have allowed this

freedom. However, children cadavers are not as common as adult cadavers and, as

mentioned in Chapter 2, family authorization is needed and sometimes difficult to obtain.

Furthermore, depending on when the death occurred, tissue transformations may have

already taken place and introduced an issue as explained in Chapter 2 for autopsy,

especially for lung tissue that changes quickly following death. In addition, no contrast

agent is used during the acquisition of the data with cadavers for obvious reasons. This

makes the images more difficult to read than the ones acquired on living patients when

contrast agents are used. As mentioned in Chapter 2, MR imaging would appear to offer

better contrast images for soft tissues, but its longer acquisition time and the different

reconstruction technique than CT make it more sensitive to patient motion artifacts and

thus severely limits its application to pediatric patients. According to the radiologists, it

represents less than 10% of pediatric patient's images. For these reasons, only CT

images were selected.

For the second step of the CT data selection, Dr Jonathan Williams, Chair of

Pediatric Radiology at Shands Teaching Hospital, reviewed all the exams selected

according to the patient's medical history and selected only those scans from patients

considered to present normal anatomy at the time of the examination. This means that

their health conditions would not drastically alter organ volume or mass. All the studies

from patients with diseases or abnormalities that could have resulted in some organ

volume or mass errors were simply eliminated.









From there, a further step was to look at the CT sets themselves. CT sets can be

composed of several image series. These series do not always have the same number of

slices and/or the same reconstruction filters. The different reconstruction filters are

usually applied to emphasize the important characteristics needed to be observed in the

CT image for the specific study ordered. In this study, the series were chosen in order to

make the segmentation easier. The adjacency of the slices within the series and the fully

imaging of the organs of interest were also checked. Partial imaging happens commonly

as a matter of keeping the radiation dose as low as possible for an exam according to the

ALARA principle. Unfortunately, all the data sets that only image partially the organs of

interest are useless for this study. Also, all the data sets with overlapping slices or gaps

between slices were removed as well, since organ volume determinations done by the

segmentation technique were based on the fact that a voxel is associated with one specific

region of interest and one specific region is fully defined by a voxel.

Segmentation Technique

When the data met all the requirements previously described, an in-house

segmentation program (Nipper 2002) developed at the University of Florida was used.

This program, called CT_Contours, is a Microsoft Windows based program written with

Interactive Data Language (IDL) version 5.5 (IDL Version 5.5, Microsoft Windows

(Win32x86) 2001). The routine called Ct Contours was mainly used to perform

segmentation and to determine organ volumes. This user-friendly program includes

several features such as the conversion of the CT data into a useful format for

segmentation, volume calculation, and 3D reconstruction.

The conversion consists in transforming the CT images from a DICOM (Digital

Communication in Medicine) format into a '.jcn' format, created by Josh Nipper, to make









the segmentation more practical. This specific function creates a file with an information

header followed by the image data set. The information header contains all the useful

information such as the data array size and dimension to assess the volume of the region

of interests. The image data set becomes a single file rather than the one-file-per-slice

technique of the DICOM format. This puts all the slices in the same order of succession

and all the parameters that identify the images and the patient are removed. As a

consequence, the handling and the segmentation of the data are easier and Health

Insurance Portability and Accountability (HIPPA) compliant.

The segmentation itself could start when the *.jcn file is loaded in the Ct Contours

routine. Many tools are available with this routine to facilitate the segmentation. As in

most medical imaging systems, the 'window' and 'level' features are also available in the

Ct Contours routine. A minimum and a maximum window value can be selected to

adjust the contrast of the image to better identify different regions and to conduct

automatic segmentation more easily. The values lying between the minimum and the

maximum are linearly interpolated between 0 and 255 (8 bits). Among the fifteen

different filters available to enhance the image quality, only two were used for this study:

the sharpen filter to increase the overall contrast of the regions and sometimes the

histogram equalization filter to differentiate tissues in manual segmentation of soft tissue.

Different segmentation schemas were run afterwards. Only when the contrast between

the region of interest and the surrounding tissues was high enough the segmentation

schema could be automatic. If the contrast was not high enough, the segmentation was

performed manually. Depending on the area to segment, the semi-automatic

thresholding, the pixel and voxel growing segmentation, or the manual freehand









segmentation were used. A Wacom pen tablet Cintiq 15X helped improving the

precision and the duration of the segmentation. The manual segmentation was a lot

easier and faster using the Cintiq 15X Grip pen direct control on the screen compared to

the use of a common computer CRT screen and mouse. When creating the segmentation

routine, Nipper (2002) created a color table, named Josh CT Contours, to label each

segmented organ region with a specific color tag that would increase its contrast with the

surrounding voxels. This same convention was used for simplicity and consistency. This

also allowed the differentiation of symmetrical organs such as lungs and kidneys.

To obtain the organ of interest volume, an estimator function, available in the

routine, calculated the volumes based upon the number of pixels assigned to a specific

color within a CT slice and the resolution information found in the *.jcn file header

information (slice thickness, pixel matrix size, and image size). A 3D reconstruction

routine included in the program, named Xvolume, allowed some reconstructions to be

performed. The visualization of the volume slice, the confirmation of the segmentation

correctness, and the organ evolution were observed.

In radiation dosimetry, the mass of the region of interest is as important as the

volume. Doses are calculated in unit of energy per unit of mass, usually mJ per kg

(mGy). Furthermore, all the anthropometric models developed until now have been

essentially based on the organ mass summarized in the ICRP Publication 23. Using the

appropriate density given by Cristy and Eckerman (1980, 1987), the mass of the organs

segmented could be determined and then compared to the ICRP data. Report 46 from the

International Commission on Radiation Units and Measurements (ICRU 1992) is more

detailed and more recent than the Cristy and Eckerman report. However, all the









anthropometric models have been developed using Cristy and Eckerman's densities, so

they were used in this present study for consistency and comparison purposes.

For each organ studied (the brain, the skull, the lungs, the liver, and the kidneys),

the selection, the segmentation, and the mass calculation differ slightly from one another

as presented further below.

Organ-specific Selection and Segmentation

Brain

Table 3-1 lists all the acquisition parameters for the selected scans for brain

segmentation. All the data were taken from Head CT exams. Cases with hydrocephalus,

severe hemorrhages, and ventriculo-peritoneal shunt were removed from further

consideration. The group included 47 children from newborns to the age of -17 months.

Thirty-one of the patients were boys (66%). The patient's mean age was 4.4 +5.14

months. The patient age distribution is presented on Fig. 3-1. Forty-seven percent of the

patients were under 3 months of age at the time of the examination. This is explained by

the high number of seizure activity early in life.

The patients were supine for all the scans. Their heads usually went first through

the scanner. The axial mode was the most commonly used acquisition mode. A tilt angle

was required sometime depending on the ordered study protocol. The slice thickness

varied from 2.5 mm to 5 mm with an average of 3.27 mm. The number of slices per

study ranged from 26 to 56 with an average of 37.57 slices per study. The display field

of view (DFOV) ranged from 130 x 130 mm2 to 200 x 200 mm2. The in-plan resolution

varies from 0.25 mm to 0.40 mm with an average value of 0.32 mm. The voxel volumes

were found to be between 0.16 mm3 and 0.73 mm3 with an average of 0.34 mm3. A

standard reconstruction filter was frequently used and preferred to the bone









reconstruction filter that was also used on occasion. The reason for this choice is that the

bone filter improves significantly the differentiation of the bone content which is not

particularly useful for this study of brain volumes.

The brain was defined as the group of subregions inside the cranium, which

includes the cerebrum, the cerebellum, the diencephalons (constituted of the thalamus

and the hypothalamus, the ventricles, and the mesencephalon), the pons, and the medulla

oblongata. The spinal cord that began caudal to the medulla oblongata was not included

in this definition. The cerebrospinal fluid (CSF) surrounding the brain was not separated

from the brain since this differentiation would not be of primary importance for the

purpose of this study. This brain definition corresponds to the definition found in the

ICRP Publication 23. It has also been called the intracranial volume by some authors

(Abbott et al. 2000; Sgouros et al. 1999).

The brain is a soft tissue organ surrounded by bone tissues, the skull. The contrast

between these two tissues is high enough to perform limited semi-automated

segmentations. The 2D and 3D pixel growing segmentation schemas were thus used.

The sharpen filter was applied to enhance the brain-skull edges. The window levels were

set between 757 and 2187 for entire segmentation of each case. Nevertheless, cranial

bones are not completely fused together for most of the children in their first year of life;

consequently, the automated growing schemas occasionally expanded the brain tissue

segmentation to the surrounding tissues outside the skull. Manual freehand segmentation

then had to be applied to remove these superfluous tissues. Some tissues had to be added

as well since they are parts of the brain, but were not included during automated









segmentation because their gray level was too different from the tissues surrounding the

brain. This usually happened with the fluid-filled ventricles.

Cristy and Eckerman (Cristy 1980; Cristy and Eckerman 1987) only defined a

generalized soft tissue density of 1.04 g.cm-3, which was used to calculate brain mass.

The ICRU Report 46 (ICRU 1992) defines a specific brain density of 1.03 g.cm-3 for a

child in his first year of life. This difference is not significant enough to introduce

important differences in mass calculation and was not taken into account.

Skull

Like the brain, head CT exams were chosen for the skull study. All the acquisition

parameters of the head CT exams selected for skull volume determination are indicated

on Table 3-1 since the exams used for the skull study were the same as the ones selected

for the brain study. However, when head CT exams were ordered for brain examination,

the skull was often not fully scanned. Therefore, these CT scans were unusable in the

case of the skull study. Consequently, there are fewer skull cases than brain cases.

Furthermore, some exams were specific cranial exams for which 3D visualization is

required afterwards. This demands a high resolution and involves very thin slice

thickness and involved image overlaps. As a consequence, these scans also had to be

removed from this study. For ease of handling, the case number for the skull volume

determination was kept corresponding to the brain case number.

The group finally selected for study included 29 children from the newborn to just

over 14 months of age. Twenty of the cases (69%) were boys. The patient's mean age

was 3.51 + 4.87 months. The patient age distribution is presented on Fig. 3-2. The









reason why there are more cases between 0 and 3 months is still the same as for the brain

study since the data are the same.

The average image characteristics used in this study are slightly different than those

used in the brain study, but this difference is not significant since only few cases have

been removed and will not be indicated here. The standard reconstruction filter was

preferred over the bone reconstruction filter when possible. The bone reconstruction filter

allows a better differentiation of intrabone tissues. The contrast between the bone tissue

and the bone marrow is thus enhanced. This would have made the segmentation more

difficult than with the standard reconstruction filter since we only want to assess the

whole skull volume in this study.

The skull normally includes 22 bones: 8 bones of the cranium and 14 bones of the

face. The skull is defined here as the bones composing the cranium and the upper face

such as parietal, frontal, temporal, occipital, sphenoid, ethmoid, and maxillary bones.

The mandible that constitutes the lower face was excluded from the skull definition here

so that we remained consistent with the work of Nipper (Nipper et al. 2002). No attempt

to dissociate the upper jaw teeth from the maxillary bone was performed. The teeth were

thus accounted for in the skull volume.

The segmentation of the skull was almost entirely automatic. The contrast was

high enough to use the 3D thresholding tools with a threshold between 1110 and 1150 for

the pixel value depending on the image quality. The sharpen filter was applied as was

done for the brain study with window levels from about 780 to 1604. The sharpen filter

helped to enhance the skull-soft tissue edges. Some manual segmentation had to be









performed to remove the bones from the mandible and from the spine as well as medical

devices like the CT table, the tubes inside the patient, or the catheters around him/her.

Cristy and Eckerman (Cristy 1980; Cristy and Eckerman 1987) defined a bone

tissue density of 1.22 g.cm-3 for the newborn and 1.4 g.cm-3 for all other ages throughout

the skeleton. These densities represent a homogeneous mixture of bone marrow and the

osseous tissues of the bone trabeculae and cortical cortex. A density of 1.61 g.cm-3 has

been defined in the ICRU Publication 46 (ICRU 1992) for the cranium of an adult only.

This density was thought to be too high for a child in his first year of life. Thus a linear

interpolation of the two Cristy and Ekerman bone densities was determined between the

newborn and the density applied here for the one-year-old and above.

Lungs

Table 3-2 lists all the acquisition parameters of the selected scans for lung

segmentation. The data were taken from Chest or Chest-Abdomen-Pelvis (CAP) CT

exams. Cases presenting expanded abscess, large amount of extra fluid, or severe

infections inside the chest were not considered. Every exam for which the lungs were

severely collapsed was removed as well because the boundaries between the lungs and

the surrounding tissues were usually barely visible and would have led to very large

uncertainties. However, exams of slightly collapsed lungs were kept as long as the

boundaries of the fully-inflated state were easy to define. The group included 20 children

between newborn and -16 months of age. Fourteen of the patients (70%) were boys. The

patient's mean age was 7.09 + 4.72 months. The patient age distribution is presented in

Fig. 3-3. There are fewer cases for this lung study than for the brain study because CT is

not the primary imaging modality used to identify pediatric chest problems.









The patients were supine and their feet went first through the scanner for almost all

these scans, Chest or Chest-Abdomen-Pelvis (CAP) exams. The helical mode was the

only acquisition mode used. No scans were performed with a tilt angle. The slice

thickness ranged from 1.25 mm to 6 mm with an average of 3.44 mm. The number of

slices per study varied from 26 to 101 with an average of 56 slices for chest and CAP

studies. CAP studies only have 65, 54 and 49 slices. The display field of view (DFOV)

ranged from 132 x 132 mm2 to 250 x 250 mm2. The in-plan resolution varied from 0.26

mm to 0.49 mm with an average of 0.34 mm. The voxel volumes are found between 0.12

mm3 and 0.91 mm3 with an average of 0.42 mm3. Usually the scans were reconstructed

with both a standard and a lung reconstruction filter. The standard reconstruction filter

was preferred in this study over the lung reconstruction filter. The lung reconstruction

filter allows a better differentiation of the contents of the lungs. Consequently, the

contrast between the air cavities, the blood vessels, and the bronchial trees that go inside

the lung cavities is enhanced. This would have made the segmentation more difficult than

with the standard reconstruction filter since we only want to assess the lung cavity

volume as a whole in this study.

The lungs were defined as the pulmonary cavity including the bronchi and the

pulmonary arteries and veins that go inside the lungs. Lungs do not have the same

volume during the entire respiratory cycle. Furthermore, a child in his first year of life

does not hold his breath when undergoing a CT examination. The CT scan can be taken

at any time during the respiration cycle and the lung volume can be influenced

accordingly. However, young children breathe either quietly when asleep or quickly but









not deeply when nervous. This negligibly influences the lung volume determination and

was not taken into account at all.

The segmentation of the lung was semi-automated. The threshold segmentation or

the 3D pixel growing segmentation schemes could not be performed since we wanted to

separate the left and right lung individually. The 2D region growing and 2D pixel

growing segmentation schemes were then selected. The sharpen filter was applied to

enhance the lung-soft tissue and lung-rib edges. The window levels were set up between

870 and 1302 for the entire segmentation of each case. However, manual segmentation

had to be completed afterwards, either to remove tissues, which were not parts of the

lungs, either to add tissues, which was parts of the normal lungs, but not included during

automated segmentation. The upper tracheobronchial tree, the lung collapsing, or the

amount of extra fluid inside the lung were cases where manual segmentation had to be

performed.

Lung tissue is defined as a mixture of blood, air, and soft tissue. Only the soft tissue

part of lung has a long-term transformation after death. A density of 0.296 g.cm-3 has

been indicated by Cristy and Eckerman (1980,1987) for all ages. ICRU Report 46 (ICRU

1992) defined the lung density as 0.26 g.cm-3. Unfortunately, this value is only for

adults. The Cristy and Eckerman's density was finally chosen since it has been used for

all the previous models.

Liver and Kidneys

Table 3-3 summarizes the acquisition parameters of the CT exams chosen for liver

and kidney segmentation. The data were mainly taken from CAP, Abdomen-Pelvis (AP),

and abdomen CT scans. One chest study was included because the liver was fully

imaged. The cases of expanded abscess or extra fluid, severe infections inside the









abdomen, and the cases of kidney or liver malfunctioning or liver transplant were not

considered. The examinations performed without contrast agent were also removed from

liver segmentation because the differentiation between the liver and the surrounding

tissues was too small to allow accurate segmentation and volume determination. The

group included 13 children ranging from newborns to -16 months. Seven of the patients

(54 %) were boys. The patient's mean age was 5.65 + 5.22 months. The patient age

distribution is presented on Fig. 3-4. For the same reasons given for the lung study, CT is

not the primary imaging modality used to identify pediatric abdomen problems.

The patients were supine and their feet went first through the scanner for almost all

these exams. Like for the lung study, the helical mode was the only acquisition mode

used and no acquisition was performed at a tilt angle. The slice thickness ranged from

2.5 mm to 5 mm with an average of 4.13 mm. The number of slices per study ranged

from 36 to 81 with an average of 53.15 slices per study. The display field of view

(DFOV) ranged from 118 x 118 mm2 to 223 x 223 mm2. The in-plan resolution varied

from 0.23 mm to 0.44 mm with an average of 0.34 mm. The voxel volumes were found

between 0.20 mm3 and 0.79 mm3 with an average of 0.49 mm3. Usually the standard and

lung reconstruction filters were used for these studies. The standard reconstruction filter

was preferred since the lung reconstruction filter was in no way helpful to segment the

liver or the kidneys.

The liver was defined as a full organ including all interior blood vessels. The

gallbladder was excluded even though it was included in most of the autopsy studies

(Urata et al. 1995). The liver shape defers from individual to individual as well as from

time to time. However, its volume does not largely vary over time.









The kidneys were defined as a full organ including blood vessels, but excluding the

urethra. The kidney is one of the only organs in the body that has a very constant shape

over time. Its volume is defined as its interior blood supply and the organ parenchyma.

The segmentation of the liver was entirely manual. The kidney segmentation was

semi-automatic when enough contrast agent was injected into the patient. However,

when the contrast was not high enough to use any of the semi-automatic segmentation

tools, manual segmentation was performed. The sharpen filter was applied to enhance

the contrast, but the histogram equalization filter was used as well to differentiate liver or

kidney tissues from the other surrounding soft tissues. The window levels were set

between 981 and 1570 for the liver segmentation and between 981 and 1302 for the

kidneys segmentation.

Cristy and Eckerman (Cristy 1980; Cristy and Eckerman 1987) only defined a soft

tissue density of 1.04 g.cm-3, while ICRU Report 46 (ICRU 1992) defined a density of

1.05 g.cm-3 for the liver of a 1-year-old and a density of 1.04 g.cm-3 for a the kidney of a

2-year-old kidney. These differences are not significant and the Cristy and Eckerman

density was thus used for both the organs for reasons of consistency discussed previously.

Graphical Analysis

None of the organs is expected to have a linear growth curve over the first year of

life. A graphical data analysis was performed with the software SigmaPlot 2001

(Brannan 2001). The organ volume was plotted against patient's age. Best-fit curves

were calculated to represent the changes in organ volume that accompany increasing age,

creating a model of normal organ growth that is useful for reference purposes. When

enough cases were available, a gender differentiation was also performed to emphasize

the difference between males and females. While it would have been highly desirable to









obtain patient total mass as the independent parameter for organ volume or mass, this

data was not readily available retrospectively, and thus organ volume as a function of

patient age is presented only.

The number of cases used for this study is low to state any kind of average of organ

volume for every month of age. However, to be able to make a comparison with the

ICRP Publication 23 data, some age intervals definitions were made. Four or five

intervals were defined to match the ICRP Publication 23 intervals: newborn, 0 to 3

months, 3 to 6 months, and either 6 to 12 months and 1 to 2 years, or 6 to 9 months, 9 to

12 months, and 1 to 2 years. As nothing was specified in ICRP Publication 23 about the

definition of the newborn, it was decided to define the newborn as babies undergoing CT

procedures in their first day-of-life. The 0-to-3-month interval includes babies from 1

day up to 3 month old. The 3-to-6-month interval starts at 3 months and 1 day and ends

at 6 month, and so on. All the results will be presented in the next chapter.














Table 3-1. Head CT exam characteristics for brain and skull studies.


Case #
Reconstruction
for Brain
B-001 Standard
B-002 Standard
B-003 Standard
B-004 Standard
B-005 Standard
B-006 Standard
B-007 Standard
B-008 Standard
B-009 Standard
B-010 Standard
B-011 Standard
B-012 C30s
B-013 Standard
B-014 Standard
B-015 Standard
B-016 Standard
B-017 Standard
B-018 Standard
B-019 Standard
B-020 Standard
B-021 Standard
B-022 Standard
B-023 Standard
B-024 Standard
B-025 H30s
B-026 Standard
B-027 Standard
B-028 Standard
B-029 H40s


Tilt Angle o Slice Thickness In-Plane Voxel Volume Scanner Case #
() # of Slices ( ) Resoluto ( ) (mm3) DFOV (mm) Mode Patient Position apeor Skul
() (mm) Resolution (mm) (mm) Type" for Skull


0
0
0
0
0
0
13
0
0
0
-15
0
-17.5
0
0
0
0
0
0
-12
0
0
19
0
0
-13
0
-6.5
0


5
2.5
2.5
2.5
3.75
2.5
2.5
2.5
2.5
2.5
2.5
4
5
2.5
3.75
2.5
2.5
2.5
2.5
3.75
2.5
2.5
2.5
2.5
4.5
3.75
3.75
3.75
3


0.38
0.27
0.27
0.26
0.29
0.35
0.26
0.28
0.31
0.29
0.28
0.28
0.33
0.35
0.39
0.30
0.35
0.39
0.31
0.31
0.30
0.25
0.31
0.28
0.30
0.33
0.28
0.33
0.35


196x196 Axial Head first Supine GE N/A
140x140 Axial Head first Supine GE S-002
140x140 Axial Head first Supine GE S-003
135x135 Axial Head first Supine GE N/A
150x150 Axial Head first Supine GE S-005
178x178 Axial Head first Supine GE S-006
132x132 Axial Head first Supine GE S-007
142x142 Axial Head first Supine GE S-008
161x161 Axial Head first Supine GE S-009
151x151 Axial Head first Supine GE S-010
142x142 Axial Head first Supine GE S-011
142x142 Helical Head first Supine Siemens N/A
168x168 Axial Head first Supine GE S-013
180x180 Helical Head first Supine GE N/A
200x200 Axial Head first Supine GE N/A
156x156 Axial Head first Supine GE N/A
180x180 Axial Head first Supine GE N/A
200x200 Axial Head first Supine GE N/A
160x160 Axial Head first Supine GE S-019
160x160 Axial Head first Supine GE N/A
152x152 Axial Foot First Supine GE S-021
130x130 Helical Head first Supine GE S-022
160x160 Axial Head first Supine GE N/A
144x144 Axial Head first Supine GE N/A
152x152 Helical Head first Supine Siemens S-025
169x169 Axial Head first Supine GE S-026
145x145 Axial Head first Supine GE S-027
167x167 Axial Head first Supine GE S-028
180x180 Helical Head first Supine Siemens S-029














Table 3-1. Continued.
Case #
S Reconstruction
for Brain
B-030 H40s
B-031 H40s
B-032 Standard
B-033 H30s
B-034 C30s
B-035 Standard
B-036 H30s
B-037 H30s
B-038 H30s
B-039 C30s
B-040 H40s
B-041 C30s
B-042 C30s
B-043 H40s
B-044 H40s
B-045 H40s
B-046 H40s
B-047 H40s


Tilt Angle # of Slices Slice Thickness In-Plane
(0) (mm) Resolution (mm)
0 28 4.5 0.29
0 36 3 0.33
0 36 2.5 0.27
0 28 4.5 0.30
0 28 4.5 0.35
0 36 3.75 0.39
0 28 4.5 0.35
0 32 4.5 0.35
0 42 3 0.29
0 47 3 0.34
0 30 3 0.27
0 36 3 0.25
0.5 32 4.5 0.35
0 32 4.5 0.39
0 42 3 0.40
0 36 3 0.26
0 42 3 0.38
0 36 3 0.34


Voxel Volume
(mm3)
0.38
0.33
0.19
0.41
0.56
0.57
0.56
0.56
0.26
0.35
0.22
0.19
0.56
0.69
0.48
0.21
0.43
0.34


DFOV
(mm)
148x148
170x170
140x140
155x155
180x180
200x200
180x180
180x180
150x150
176x176
140x140
130x130
180x180
200x200
204x204
134x134
194x194
172x172


a GE, GE light Speed 4-slice scanner; Siemens, Siemens-Somaton sensation 16 scanner
b N/A, CT head exams cannot be used for skull volume calculation


Mode

Axial
Axial
Helical
Axial
Axial
Axial
Axial
Axial
Axial
Helical
Axial
Axial
Axial
Axial
Axial
Axial
Axial
Axial


Patient Position

Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine
Head first Supine


Scanner
Typea
Siemens
Siemens
GE
Siemens
Siemens
GE
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens


Case # for
Skullb
S-030
S-031
S-032
S-033
S-034
N/A
N/A
S-037
S-038
N/A
N/A
S-041
S-042
S-043
N/A
N/A
N/A
N/A














Table 3-2. CT exam characteristics for lung study.


Case # for
Lung
LU-001
LU-002
LU-003
LU-004
LU-005
LU-006
LU-007
LU-008
LU-009
LU-010
LU-011
LU-012
LU-013
LU-014
LU-015
LU-016
LU-017
LU-018
LU-019
LU-020


Study
type"
CAP
Chest
Chest
CAP
Chest
Chest
Chest
Chest
Chest
Chest
Chest
Chest
Chest
CAP
Chest
Chest
Chest
Chest
Chest
Chest


Reconstruction

B30f
B30f
B30f
B30f
Standard
Standard
Standard
B30f
Standard
Standard
Standard
B30f
Standard
B30f
B30f
B30f
B30f
B30f
B30f
B30f


# of
Slices
49
52
26
54
101
73
57
79
45
63
41
56
49
65
34
68
83
51
36
35


Slice Thickness
(mm)
5
2
5
5
1.25
2.5
2.5
2
3.75
2.5
2.5
3
3.75
5
6
2
2
3
5
5


In-Plane
Resolution (mm)
0.30
0.27
0.26
0.31
0.31
0.49
0.36
0.32
0.45
0.39
0.40
0.33
0.41
0.32
0.39
0.26
0.31
0.35
0.35
0.30


Voxel Volume DFOV
V l V e D V Mode Patient Position
(mm3) (mm)
0.45 154x154 Helical Foot First Supine
0.15 140x140 Helical Head First Supine
0.34 134x134 Helical Foot First Supine
0.48 158x158 Helical Foot First Supine
0.12 160x160 Helical Foot First Supine
0.60 250x250 Helical Foot First Supine
0.33 186x186 Helical Foot First Supine
0.21 166x166 Helical Foot First Supine
0.74 228x228 Helical Foot First Supine
0.39 201x201 Helical Foot First Supine
0.39 203x203 Helical Foot First Supine
0.33 170x170 Helical Foot First Supine
0.64 211x211 Helical Foot First Supine
0.51 164x164 Helical Foot First Supine
0.91 199x199 Helical Foot First Supine
0.13 132x132 Helical Head First Supine
0.20 161x161 Helical Foot First Supine
0.37 180x180 Helical Foot First Supine
0.62 180x180 Helical Foot First Supine
0.45 154x154 Helical Foot First Supine


Scanner Typeb

Siemens
Siemens
Siemens
Siemens
GE
GE
GE
Siemens
GE
GE
GE
Siemens
GE
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens
Siemens


a CAP, chest-Abdomen-Pelvis Study
b GE, GE light Speed 4-slice scanner; Siemens, Siemens-Somaton sensation 16 scanner













Table 3-3. CT exam characteristics for liver and kidney studies.
Case #
for Study # of Slice Thickness In-Plane Voxel Volume DFOV Scanner Case # for
for a Reconstruction 3 Mode Patient Position b K c
aer Type Slices (mm) Resolution (mm) (mm3) (mm) Type Kidney
Liver
N/A AP Standard 81 2.5 0.35 0.31 180x180 Helical FootFirst Supine GE K-001
LI002 Abdomen Standard 53 2.5 0.35 0.31 180x180 Helical Foot First Supine GE N/A
LI003 AP Standard 69 3.75 0.44 0.71 223x223 Helical Foot First Supine GE K-003
LI004 AP B30f 44 5 0.40 0.79 204x204 Helical Head First Supine Siemens K-004
LI005 AP Standard 42 3.75 0.23 0.20 118x118 Helical Foot First Supine GE K-005
LI006 AP Standard 68 3.75 0.43 0.70 221x221 Helical Foot First Supine GE K-006
LI007 AP Standard 40 2.5 0.31 0.24 160x160 Helical Foot First Supine GE K-007
LI008 CAP B30f 49 5 0.30 0.45 154x154 Helical Foot First Supine Siemens K-008
LI009 CAP B30f 54 5 0.31 0.48 158x158 Helical Foot First Supine Siemens K-009
LIO10 CAP B30f 65 5 0.32 0.51 164x164 Helical Foot First Supine Siemens N/A
LIO11 AP B30f 40 5 0.30 0.44 152x152 Helical Foot First Supine Siemens K-011
N/A AP B30f 50 5 0.35 0.62 180x180 Helical Foot First Supine Siemens K-012
LI013 Chest B30f 36 5 0.35 0.62 180x180 Helical Foot First Supine Siemens N/A
a AP, Abdomen-Pelvis Study; CAP, chest-Abdomen-Pelvis Study
b GE, GE light Speed 4-slice scanner; Siemens, Siemens-Somaton sensation 16 scanner
c N/A, CT exams cannot be used for liver or kidney volume calculation












16
Female
14

S12 Male
10

8

6


I w- I---I--I-------------------
4i l




o ,> l ,l l, ,

Age (months)


Figure 3-1. Patient distribution of the brain study by age and sex.











12


10 Female

Male
8


6





2-



II, N b ------ C) > N g (Moths
o llI ,,,, ,I



Age (Months)


Figure 3-2. Patient distribution of the skull study by age and sex.














* Female


* Male


N 2- P ,\ t
Q N' rvf V 6 e


3




2 -




1




0
Q5


Figure 3-3. Patient distribution of the lung study by age and sex.


(months)












3


5

1


0


* Female

* Male


SI II I I
SN P rb v ) t'Age (months) N N N\
Age (months)


Figure 3-4. Patient distribution of the liver and kidney studies by age and sex.














CHAPTER 4
RESULTS AND DISCUSSION

The results and discussion will be divided in three parts: (1) the establishment of

the organ-specific growth curves, (2) the comparison with references and already

published data if any, and (3) the characterization of the stylized and tomographic

models.

Organ Growth Curves

In this first step of the establishment of the organ-specific growth curves, the

general trend of organ evolution was determined and will be detailed later in this chapter

for the brain, the skull, and the lungs. Three-parameter hyperbola and three-parameter

power equations were found to be the best-fit curve empirical equations for the different

organ growth curves presented separately hereafter. The three-parameter hyperbola is

described by the equation:

a AGE
VOL = V AG (4-1)
b+AGE

In Eq. (4-1) VOL represents the organ volume in cm3 at a given age, Vo is the

supposed organ volume at birth, a and b are two coefficients that will characterize the

growth curve, and AGE is the age of the child in months. The three-parameter power

equation is described by:

VOL = Vo +c.(AGE)d. (4-2)









In Eq. (4-2) VOL represents the organ volume in cm3 at a given age, Vo is the

supposed organ volume at birth, c and d are two coefficients that will characterize the

growth curve, and AGE the age of the child in months.

When it was possible, a gender distinction was made. This data differentiation

usually demonstrates the difference between girls and boys in a manner similar to that for

height, weight, and head circumference (Clinical Growth Charts 2001). All the

specificities of each organ studied are detailed hereafter.

Brain

Table 4-1 lists the characteristics of each patient selected for the brain study

including gender, age, and brain volume. Forty-seven brains have been segmented from

newborn to -17 months of age. The smallest brain is only 258.85 cm3 for a 0.3 month-

old child and the largest is 1232.03 cm3 for 14.29 month-old. There are important

differences between individuals even at the same age. Fig. 4-1 was created from the data

of Table 4-1. The brain volume is plotted against the age of the patient. This represents a

model of normal brain growth over the first year-of-life regardless the gender. The data

spread observed in the first months of life can reflect the difference between babies born

prematurely or at normal gestation. Unfortunately, limited access to the medical records

did not allow identification of those born prematurely. Nonetheless, premature birth does

not make any difference after a couple of months (Ounsted et al. 1982). The principal

differences essentially come from family characteristics, like parents' height, and the

severity of birth defects and illnesses. However, patients in serious conditions have been

withdrawn from this study. Only patients with nominally healthy condition remain. The

main factors of difference in this study are finally the premature birth in the first few

months of life and the parents' morphometric characteristics.









The general trend is a rapid growth in the first weeks of life and then the rate slows

with increasing age. The most appropriate equation to fit the curve has been found to be

described by the three-parameter hyperbola:

1092.77 -AGE
VOL = 374.04 + 92 GE(4-3)
8.762 + AGE

The coefficient of determination (R2) is 0.84. Sgouros et al. (1999) did a similar

study for patients from age 7 days to 15 years using magnetic resonance imaging. When

comparing the brain volume (called intracranial volume in Sgouros et al. (1999) study)

from the patients under one year of age (13 cases) of this study with the brain volume of

our study, the results are similar to those shown in Fig. 4-2. Abbott et al. (2000) did not

specify their results as only graphical data were published. Furthermore, they determined

the age of the children from conception to the day of the examination. Therefore, the

comparison would not have been as relevant.

Fig 4-3 shows the difference in normal brain development between males and

females. Male brain volumes best-fit curve is constantly above the female fit curve. This

agrees with head circumference charts, in which a similar difference between the two

genders has been established (Clinical Growth Charts 2001). The most appropriate

equations to fit these curves have also been found to be three-parameter hyperbolas. For

females, the equation has been defined as:

912.46. AGE
VOL = 288.27+ 92 (4-4)
4.67 + AGE

The coefficient of determination (R2) is 0.77. For males, the equation has been

defined as:

1396.8.AGE
VOL = 394.2 + (4-5)
13.6+AGE









The coefficient of determination (R2) is 0.90. In most of the published data, the

curves for males and females differ from one another but the evolution is parallel (Abbott

et al. 2000; Sgouros et al. 1999). Here this is not the case. A possible explanation is the

small number of female data (only 15 cases).

Table 4-2 summarizes the average volumes over five different periods of time in

the first year-of-life. It also indicates the minimum and the maximum brain volumes for

each period as well as the standard deviation and the percentage difference relative to the

mean (column 2) between the maximum and minimum volumes. For a newborn, the

mean brain volume is 406.83 cm3. The smallest brain volume is 331.57 cm3 and the

largest is 518.92 cm3. This volume reaches 451.08 cm3 during the first three months of

life. It increases to 697.51 cm3 on average between the ages of 3 and 6 months to

approach 940 cm3 between 6 and 12 months of age. This average attains 1010.24 cm3

between 1 and 2 years of age. This represents a 250 % brain volume increase over the

first year-of-life.

Skull

Table 4-3 lists each patient selected for the skull study including gender, age, and

skull volume. There are large differences between individuals even at the same age.

Twenty-nine skulls have been segmented from the newborn to -14 months. The smallest

skull is only 58.17 cm3 for a 0.3 month-old and the largest is 344.07 cm3 for a 14.29

month-old. These results are in correlation with the corresponding results for brain

volume. Fig. 4-4 was created from the data of Table 4-3. The skull volume is plotted

against the age of the patient. This represents a model of normal skull growth over the

first year-of-life regardless the gender.









The general trend is the same as the brain volume growth, since the brain can also

extend if the skull extends. The most appropriate equation to fit the curve has been found

to be a three-parameter hyperbola described by:

441.24 AGE
VOL = 81.74 + (4-6)
14.38 +AGE

Equation (4-6) is represented on Fig. 4-4. The coefficient of determination (R2) is

0.91. Unfortunately there are no published data with which to make comparisons.

Medically, the skull volume is of little clinical interest. The most important skull

characteristics monitored are usually the intracranial dimensions (Waitzman et al. 1992a;

Waitzman et al. 1992b) to sometimes determine the intracranial volume. These

dimensions are critical for most of the craniofacial surgeries like for craniosynostosis.

The volume of the skull bone (Buda et al. 1975) is rarely measured.

The difference in normal skull development between males and females is not

presented here, because the number of cases for female children is not high enough (9) to

allow any comparison with the male children (20).

Table 4-4 summarizes the average volumes over five different periods of time in

the first year of life. It also indicates the minimum and the maximum skull volume for

each period as well as the standard deviation and the percentage difference relative to the

mean (column 2) between the maximum and minimum volume. For a newborn, the mean

skull volume is 87.56 cm3. The smallest skull volume is 78.79 cm3 and the largest is

101.43 cm3. This volume reaches 102.89 cm3 during the first three months of life. It

increases to 154.89 cm3 on average between the age of 3 and 6 months to attain 232.6

cm3 between 6 and 12 months of age. This average approaches 300 cm3 between 1 and 2

years of age. This represents a 340 % skull volume increase over the first year-of-life.









If the skull volume may not have an important interest out of the radiation

dosimetry field, the segmentation allowed 3D reconstruction that shows another view of

the skull evolution. The growth of the different bones of the skull until they all fused

together is well imaged with this reconstruction. If the fontanels are large at birth, they

almost disappear within the first few months of life as shown on Fig. 4-5.

Lungs

Table 4-5 lists each patient selected for the lung study including gender, age, total

lung volume, left and right lung volume, and the difference between the left and right

lung volume. The right lung is compared to the left lung. Twenty pairs of lungs have

been segmented from the newborn to -16 months. The smallest pair of lungs is only

163.12 cm3 for a 0.23-month-old and the largest is 621.13 cm3 for 13.6-month-old. There

are large differences between individuals even at the same age. Fig. 4-6 was created

from the data of Table 4-5. The total lung volume is plotted against the age of the

patient. This represents a model of normal lung growth over the first year-of-life

regardless the gender. The general trend is different from the brain and the skull studies.

The lungs display a nearly constant growth rate over the first year-of-life. The most

appropriate equation to fit the curve has been found to be a three-parameter power

equation described below and represented on the Fig. 4-6.

VOL = 167.42 +48.62.(AGE)068 (4-7)

The coefficient of determination (R2) is 0.55. Unfortunately there is no published

data with which to make comparisons for our study. The only published data for lung

volume calculations are essentially performed for elderly people with lung cancer or

chronic diseases.









The difference in normal lung development between males and females is not

presented in this study because there are not enough cases for both genders (6 female

cases and 14 male cases). However, the left and right lung development is indicated in

Fig. 4-7, as well as the difference between their volumes in Fig. 4-8. The left lung is

generally smaller than the right lung. The difference between both lungs can greatly vary

between individuals even in the same age category. The most appropriate equations to fit

these curves have also been found to be three-parameter power equations. For the left

lung, the equation has been defined as:

VOL = 67.91+ 28.22.(AGE)062 (4-8)

The coefficient of determination (R2) is 0.53. For the right lung, the equation has

been defined as:

VOL = 99.05 + 21. (AGE) 74 (4-9)

The coefficient of determination (R2) is 0.55. Nonetheless, no difference in

evolution over the time can be set as shown in Fig. 4-8 and in Fig. 4-9. The largest

difference between the left and the right lung is 37 % for a 1.97-month-old and the

smallest is 1 % for a 8.71-month-old. The overall average difference is 16 % and gives

an average left lung volume 45.6 % of the total lung volume (TLV). This agrees with the

left lung accounting for between 45.5% and 49.1% of the TLV in different studies

already performed on adults using different methods (Denison et al. 1986; Horsefield

1967; Pierce et al. 1980). The constant average difference between left and right lung

throughout life is emphasized in this study.

In one case the right lung is smaller than the left lung. This situation is referred to

as a volume loss. This can be occurring due to a normal congenital variation, a









pneumonia, which is a pulmonary infection, or even an atelectasis, a bronchi collapsing

due to mucus. This is not uncommon. In this particular case, it was not considered

severe enough to be removed from our study. The case where the left lung is

significantly smaller than the right lung can also reflect the volume loss problem that

usually disappears over time.

Table 4-6 summarizes the average volumes over five different periods of time in

the first year-of-life. It also indicates for each period as well as the standard deviation and

the percentage difference relative to the mean (column 2) between the maximum and

minimum volume. The total lung volume is 200 cm3 on average during the first three

months-of-life. It increases to 337.28 cm3 between the age of 3 and 6 months to attain

390.5 cm3 between 6 and 9 months of age. This average approaches 357 cm3 between 9

and 12 months of age to reach 516.38 cm3 between 1 and 2 years of age. The average

total lung volume for the 9-to-12-month-old period is smaller than the 6-to-9-month-old

period. This may be due to the small number of cases for both periods. The cases may

not truly represent an average population even though the volume difference over these

periods is small. The volume increases by 260 % over the first year-of-life.

Liver

Table 4-7 lists each patient selected for the liver study including gender, age, and

the liver volume. Eleven livers were segmented from the newborn to -16 months. The

smallest liver is only 148.8 cm3 for a 0.23 month-old and the biggest is 489.33 cm3 for

12.39 month-old. There are large differences between individuals even at the same age.

However, there are not enough cases to conduct a complete study like the brain, the skull,

or the lung studies presented previously.









Kidneys

Table 4-8 lists each patient selected for the kidney study including gender, age,

total kidney volume, left and right kidney volume, and the difference between the left and

right lung volume. Ten pairs of kidneys were segmented from the newborn to -12

months. The smallest pair of kidneys is only 21.46 cm3 for a 0.03 month-old and the

largest is 95.52 cm3 for a 12.39 month-old. There are large differences between

individuals even at the same age. In general, the left kidney is slightly larger than the

right kidney. This statement agrees with the ICRP Publication 23, although the

quantification of this difference is not indicated. In this study, the largest difference is

20% for the 12.39-month-old and the smallest is 1 % for a 2.17-month-old. The overall

average difference is 7%. Fig. 4-10 represents this difference. Brown et al. (2001)

mentioned a 4% different in their study of 8 pediatric kidneys. In two cases the right

kidney is larger than the left kidney. There is no explanation for this fact. However, this

study has too few cases to go further than giving the volume and mass results. Any

statement would not be speculative.

Comparisons with ICRP Reference Values

In the second part of the study, some comparisons with the ICRP Publication 23

were performed because this report has always been the reference to build anthropometric

models for radiation dosimetry. In this report the main organ references are organ mass

as differentiated by gender. As a consequence, before making some comparisons with

the ICRP Publication 23 values, organ masses were calculated using the CT-based organ

volumes and the Cristy and Eckerman's densities indicated in Chapter 3 (Cristy 1980;

Cristy and Eckerman 1987). The average over both genders was also calculated for the









ICRP Publication 23 data to allow the comparison between CT-based and ICRP-report

organ masses. The CT-based values were too few to allow a gender comparison.

Table 4-9 lists the CT-based brain masses and the ICRP 23 brain masses for five

period of growth. Fig. 4-11 represents graphically the difference between the two data

sets. The CT-based brain mass averages are larger than the ICRP averages one for each

period. Table 4-9 also indicates the percentage difference between the CT-based brain

masses and the ICRP 23 brain masses for each period. These differences range from

10.7% for the 0-to-3-month-old period to 28% for the 3-to-6-month-old period.

The ICRP publication 23 does not give any specific skull masses throughout the

life time. The total weight for the skeleton is given on a graph. Specific bones are

characterized by a percentage of this total skeleton weight for newborn and adult only.

However, the average masses of the skull over 5 periods from birth to 12 months were

calculated. Table 4-10 summarizes these CT-based skull masses.

Table 4-11 summarizes the CT-based total lung masses and the ICRP total lung

masses for five periods. Fig. 4-12 represents graphically the difference between the two

data sets. The CT-based lung mass averages are smaller than the ICRP averages one for

each period. Table 4-11 also indicates the percentage difference between the CT-based

total lung masses and the ICRP 23 total lung volume for each period. These differences

go from -26% for the 9-to-12-month-old period to 4% for the 3-to-6-month-old period.

No left and right lung masses are specified in the ICRP Publication 23.

Table 4-7 summarizes the CT-based liver masses and Table 4-8 the CT-based

kidney masses. However, for the same reason enounced previously no comparison can

be made. Too few cases are available.









Over all for mass comparison, a debatable issue can be enounced. The ICRP

publication 23 gathers a lot more cases per period than this present study. However,

these differences can be attributed to a combination of factors such as (1) overall

increases in individual size and stature of the past several decades, (2) problems with

tissue swelling or shrinkage post mortem, and (3) differences in techniques used for brain

volume and/or mass determination like it has already been stated in some publication

(Abbott et al. 2000; Sgouros et al. 1999).

Characterization of Stylized and Tomographic Models

The third and last part of the study was the characterization of the existing

anthropomorphic models using the growth curves established earlier in this chapter.

Only organ volumes were compared since organ mass comparisons would not have

brought different results. The same densities were used in the development of the

anthropometric models and the growth curves.

The differences between the model organ volume and the growth curves values

were stated per organ and for the entire model. For each organ the relative difference

between the model and the growth curve values was calculated. For the entire model, the

difference took into account the average of the absolute differences for the brain, the

skull and the TVL. The equation was defined by:


average model Aorg (0-10)
n

In Eq. (4-10) average model represents the average difference between the model and

the growth curve established above. Aorgan is the relative difference between the model









and the growth curve values. n is equal to the number of organs used to calculate the

average, n=3 in this study.

Stylized Models

To build the anthropomorphic stylized model Cristy and Eckerman followed the

references stated in the ICRP Publication 23. The brain, the skull, and the lung volumes

they used are summarized in Table 4-12 for the newborn model and for the one-year-old

model. When Bouchet et al. (1999) made a new head model, the ICRP Publication 23

was still the preferred reference source for a manner of consistency. However, the brain

volume has been increased by 10% compared to the Cristy and Eckerman's brain model

for the newborn, giving a new volume of 371.9 cm3. This volume became 1.4% smaller

for the one-year-old model, going from 850 cm3 to 838.4 cm3. Even with an augmented

brain volume, Bouchet et al. model is still small compared to the average values obtained

in this study (406.83 cm3). However, this new volume (371.9 cm3) is only 1% smaller

than the newborn value determined by Eq. (4-1) defined previously in this study (374.04

cm3). Unfortunately, the new one-year-old brain volume is still very low. The volume

defined by Eq. (4-1) above is 1005.7 cm3 for a 12-month-old child when for the one-year-

old stylized model the brain volume is only 838.4 cm3 using the new head model of

Bouchet et al..

Cristy and Eckerman's skull model includes the mandible because the facial bones

were undifferentiated, unlike Bouchet et al. stylized head model and the skull defined in

this study. The skull volume has been increased by 50 % in Bouchet et al. (1999) new

head model compared to the Cristy and Ekerman's skull model for the newborn, giving a

new volume of 83.5 cm3. This volume became 56 % bigger for the one-year-old model,

going from 161.8 cm3 to 252.8 cm3. The newborn skull volume from Bouchet et al.









appears only 5% smaller compared to the average values obtained in this study (87.56

cm3). However, this new volume (83.5 cm3) is less than 1% bigger than the newborn

value that is determined by Eq. (4-6) defined previously in this study (81.74 cm3). The

new one-year-old skull volume is also within 0.5% difference with the value defined by

Eq. (4-6). The volume defined by Eq. (4-6) above is 282.46 cm3 for a 12-month-old child

when for the one-year-old stylized model the skull volume is 252.8 cm3 using the new

head model from Bouchet et al..

The lung volume for the newborn stylized model is 171 cm3, the left lung volume is

79.1 cm3 and the right lung volume is 91.9 cm3. This represents a 14% difference

between the left and right lung volume. The total volume is about 2% bigger than the one

given by Eq. (4-7) established previously (167.42 cm3). However, the left lung volume is

16% bigger and the right one is 7% smaller than the results provided by Eq. (4-8) and Eq.

(4-9) (67.91 cm3 for the left lung and 99.05 cm3 for the right one). The lung volume for

the 1-year-old stylized model is 484 cm3, the left lung volume is 225 cm3 and the right

one is 259 cm3. This represents a 13% difference between the left and right lung volume.

The total volume as well as the left and right lung volume is about 12% bigger than the

one given by Eqs. (4-7), (4-8), and (4-9) (430.85 cm3, 199.63 cm3, and 231.12 cm3,

respectively).

Tomographic Models

The brain, the skull, and the lung volumes found for the UF-newborn (6-day-old

child) female, the UF-2-month, and the UF-9-month male tomographic model are also

summarized in Table 4-12 (Nipper et al. 2002).









The UF-newborn brain is smaller than the stylized model and 22 % smaller than the

value defined by Eq. (4-1) (398.43 cm3). However, ifEq. (4-2) is used, the brain volume

for a 6-day-old female should approach 325.74 cm3, that only represents a 5% difference.

Being aware that the female brain volume evolution in function of the age over the first

year of life was established with a small number of cases compared to the male and the

undifferentiated brain volume growth curve, this result can still be debatable. Its skull

has a volume of 97.34 cm3 (Nipper et al. 2002). This value is 10 % bigger than the

stylized model and 9 % bigger than the value defined by Eq. (4-6) (87.79 cm3). Its total

lung with a volume of 122.83 cm3, the left lung reaches 56.5 cm3, when the right one

reaches 66.33 cm3 to make a 15% difference. This pair of lungs is not the cadaver used

to build the model. It comes from a living patient who had approximately the same

characteristics than the cadaver who had largely collapsed lungs. It appears smaller than

the stylized model lungs and about 30% below the curve defined by the Eqs. (4-7), (4-8),

and (4-9) (183.69 cm3, 78.31 cm3, and 105.43 cm3, respectively).

The UF-2-month male tomographic model has been found to have a brain volume

of 535.97 cm3 (Nipper et al. 2002). Eq. (4-1) gives a brain volume of 577.16 cm3 for a 2-

month-old child and Eq. (4-3) 573.28 cm3. This represents only about 7 % difference for

both equations. Its skull has a volume of 166.98 cm3 (Nipper et al. 2002). Eq. (4-6) for a

2-month-old child gives a skull volume of 135.62 cm3. This represents about 23%

difference. Its total lung volume was determined at 156.11 cm3. The left lung was 74.53

cm3 and the right 80 cm3. This represents a 7% difference between the left and right

lung. Eq. (4-7) gives a total lung volume of 245.32 cm3 for a 2-month-old child, Eq. (4-

8) gives a left lung volume of 111.28 cm3, Eq. (4-9) 134.12 cm3 for the right lung









volume. This represents a difference over 32% for the three equations. However, the CT

data used to build this model came from a 6-month-old male cadaver. Eq. (4-1) gives an

818.25-cm3 brain volume for a 6-month-old child and Eq. (4-3) 821.79 cm3. This

represents about 1.5 times the actual child brain volume determined by segmentation.

Eq. (4-6) gives a 211.64-cm3 skull volume for a 6-month-old child. This represents more

than 1.25 times the actual child skull volume determined by segmentation. Eq. (4-7)

gives a 331.84-cm3 total lung volume for a 6-month-old child, Eq. (4-8) gives 153.92 cm3

for the left lung volume and Eq. (4-9) 178.13 cm3 for the right lung volume. This

represents about 2 times the actual child lung volume determined by segmentation. This

child was born highly premature and was very sick. His height and weight was checked

and better corresponded to an average 2-month-old child than to a 6-month-old child

according to the growth charts (Clinical Growth Charts 2001).

A 9-month-old male model has recently been built here at UF. Its brain volume

approaches 948 cm3. However, Eq. (4-1) gives a 927.8-cm3 brain volume for a 9-month-

old child when Eq. (4-3) gives 950.45 cm3. The difference is then 3% for the general

trend and 1% for the male trend. His skull volume approaches 323 cm3. Eq. (4-6) gives

a 251.59-cm3 brain volume for a 9-month-old child. The difference falls within 22 %. Its

total lung volume approaches 335 cm3, its left lung 152 cm3, and its right lung 183 cm3.

Eq. (4-7) gives a 384.04-cm3 total lung volume for a 9-month-old child when Eq. (4-8)

gives 178.11 cm3 for the left lung volume and Eq. (4-9) gives 205.8 cm3 for the right lung

volume. The difference is 11% for the right lung and 15% for the left lung.









From the results presented in Table 4-12 stylized and tomographic models were

compared to the curves established in this study. The stylized models with Bouchet et al.

head model have been observed to nicely represent the average population in terms of

volume even though these models are very simplistic.

The tomographic models trend to behave differently. The UF-newborn model is

22% smaller than average over the organs used. The lungs were imported from a living

patient. They are particularly small, even though their match with the body is visually

appropriate. The whole-body model is simply small even if the patient had an average

weight for a newborn (3.84 kg) according to the weight CDC growth chart (Clinical

Growth Charts 2001). Post mortem changes may already have taken place.

The UF-2-month model was definitively not equivalent to a 6-month-old average

baby. The age scaling done according to the total weight and height of the patient

appears appropriate, although the model still has small lungs. The brain volume appears

in the range of a 2-month-old child, but for the skull and the lungs, it does not appear

appropriate. The lungs are smaller than the newborn lung volume determined by the

growth curve (Eq. (4-7)). The skull would match a 4-month-old skull according to the

growth curve (Eq. (4-6)). With more organ comparison it would able to have more

detailed conclusions. However, it can be stated that this patient was sick and weight and

height can be a better indicator of child development than age.

The 9-month-old model is only about 14% different on average for the whole-body.

The differences with the growth curves appear smaller than the two other tomographic

models. One explanation may be the fact that this model has been developed from data






62


of a living patient and not a cadaver like the two other models. However, the reasons for

performing the full CT scan on this particular patient were not determined.











Table 4-1. Sex, age, brain volume for the 47 children used for the brain study.
Brain Brain
Case Sex Age Volume Case Sexa Age Volume
(months) (cm3) (months) (cm3)


B009 N
B011 N
B040 F
B041 F
B045 N
B024 N
B002 N
B003 N
B010 N
B016 N
B032 N
B007 N
B027 N
B012 N
B005 N
B022 N
B004 N
B019 N
B031 N
B008 F
B021 F
B033 F
B025 N
B029 F
aM, male; F, female


0.00
0.00
0.00
0.00
0.00
0.03
0.03
0.03
0.03
0.03
0.07
0.10
0.13
0.23
0.26
0.30
0.33
0.62
0.76
0.95
1.54
1.54
1.77
2.17


518.92
450.98
344.38
331.57
388.31
389.26
341.55
362.35
518.38
401.68
318.56
365.50
333.64
502.82
403.13
258.85
359.97
537.69
498.18
329.72
507.01
427.76
527.16
704.89


B030
B015
B020
B038
B047
B023
B039
B035
B013
B014
B017
B034
B028
B006
B018
B044
B026
B036
B042
B037
B043
B046
B001


2.23
2.33
2.60
3.02
3.78
4.80
4.99
6.11
6.31
6.70
7.23
7.92
8.57
9.77
10.94
11.56
12.06
12.29
13.63
13.63
14.29
14.55
16.79


474.45
702.45
658.67
557.40
662.65
669.64
895.41
1066.78
814.55
885.64
1054.27
653.30
934.11
859.78
1027.64
1163.88
857.82
978.48
1143.16
802.46
1232.03
898.34
1159.42







64


Table 4-2. Average brain volume during different period of the first year and a half of life
and its characteristics.
e I s Mean Brain STa n (m Max' Differenced
Age Intervals STD Minb (cm3)
Volume (cm3) (cm3) (%)
0 406.83 78.14 331.57 518.92 36
0-3m 451.08 124.35 258.85 704.89 63
3-6m 697.51 141.97 557.40 895.41 38
6-12m 939.99 155.84 653.30 1163.88 44
1-2y 1010.24 167.85 802.46 1232.03 35
a STD, Standard deviation
b Minimum brain volume found over the period
c Maximum brain volume found over the period
d Difference between the maximum and minimum brain volume found over the period relative to the mean











Table 4-3. Sex, age, skull volume for the 29 children used for the skull study.
Skull Skull
Case Sexa Age Volume Case Sexa Age Volume
(months) (cm3) (months) (cm3)


S009
SOll
S041
S045
S002
S003
S010
S032
S007
S027
S005
S022
S019
S031
S008


101.43
86.33
83.68
78.79
70.56
74.53
101.22
77.32
64.23
90.75
100.10
58.17
116.53
116.23
68.46


aM, male; F, female


S021
S033
S025
S029
S030
S038
S013
S034
S028
S006
S026
S042
S037
S043


1.54
1.54
1.77
2.17
2.23
3.02
6.31
7.92
8.57
9.77
12.06
13.63
13.63
14.29


101.24
136.16
159.39
180.10
131.19
154.89
240.24
259.49
222.43
207.42
241.58
332.54
279.85
344.07







66


Table 4-4. Average skull volume during different period of the first year and a half of life
and its characteristics.
Mean Skull volume Differenced
Age Intervals (c) STDa Minb (cm3) Max' (cm3)
(cm3) (%)
0 87.56 9.76 78.79 101.43 22
0-3m 102.89 35.37 58.17 180.10 68
3-6m 154.89 N/A N/A N/A N/A
6-12m 232.60 22.60 207.42 259.49 20
1-2y 299.51 47.68 241.58 344.07 30
a STD, Standard deviation
b Minimum skull volume found over the period
' Maximum skull volume found over the period
d Difference between the maximum and minimum skull volume found over the period relative to the mean





ge, lung volume for the 20 children used for the lung study.


Table 4-5. Sex, a

Case Sexa

LU005 F
LU016 M
LU002 M
LU003 F
LU011 M
LU017 M
LU020 M
LU004 M
LU006 M
LU001 M
LU008 F
LU012 F
LU010 M
LU009 M
LU013 M
LU014 M
LU007 M
LU018 F
LU015 M
LU019 F
aM, male; F, female


Age
(months)
0.03
0.26
1.81
1.97
2.96
3.78
4.60
5.36
5.72
5.88
6.31
8.64
8.71
9.23
9.33
11.17
11.79
13.60
14.95
15.67


Total Lung
Volume (cm3)
190.52
163.12
283.04
179.86
183.80
370.73
461.69
238.83
404.96
210.19
375.52
337.78
458.17
404.51
333.74
330.13
358.79
621.13
500.41
427.61


Left Lung
Volume (cm3)
80.55
67.17
143.87
69.45
80.18
169.08
209.15
113.89
179.65
102.14
185.33
138.68
227.96
182.45
156.47
162.30
154.06
297.30
237.06
188.44


Right Lung
Volume (cm3)
109.97
95.95
139.17
110.41
103.62
201.65
252.54
124.94
225.31
108.05
190.19
199.1
230.21
222.06
177.27
167.83
204.73
323.83
263.35
239.17


Difference between left
and right lung (%)
27
30
-3
37
23
16
17
9
20
5
3
30
1
18
12
3
25
8
10
21










Table 4-6. Average lung volume during different period of the first year and a half of life
and its characteristics. A) Total lung volume. B) Left lung volume. C) Right
lung volume.
A)
Age Intervals Mean Total Lung STDa Minb ( ) Max (cm) Differenced
Age Intervals Voum cSTD Mo(cm3) Max (cm)
Volume (cm3) (%)
0-3m 200.07 47.47 163.12 283.04 42
3-6m 337.28 108.42 210.19 461.69 54
6-9m 390.49 61.58 337.78 458.17 26
9-12m 356.79 34.27 330.13 404.51 18
1-2y 516.38 97.74 427.61 621.13 31
B)
Age Intervals Mean Left Lun STDa Minb (cm) Max' (cm3) Difference
volume (cm3) (%)
0-3m 88.24 31.69 67.17 143.87 53
3-6m 154.78 45.34 102.14 209.15 51
6-9m 183.99 44.66 138.68 227.96 39
9-12m 163.82 12.89 154.06 182.45 16
1-2y 240.93 54.53 188.44 297.30 37
C)
Age Intervals Mean Right Lun STD Minb (cm) Max (cm3) Differenced
Volume (cm3) (%)
0-3m 111.82 16.37 95.95 139.17 31
3-6m 182.50 63.17 108.05 252.54 57
6-9m 206.50 21.01 190.19 230.21 17
9-12m 192.97 24.92 167.83 222.06 24
1-2y 275.45 43.61 239.17 323.83 26
a STD, Standard deviation
b Minimum lung volume found over the period
Maximum lung volume found over the period
d Difference between the maximum and minimum lung volume found over the period relative to the mean







69


Table 4-7. Sex, age, liver volume for the 11 children used for the liver study.
Liver
Age iver Liver
Case Sexa Age Volume
(months) (cm3 Mass (g)
LI005 F 0.03 89.67 93.26
LI007 M 0.23 148.8 154.75
LI002 F 2.10 187.35 194.84
LIO11 M 4.70 194.51 202.29
LI009 M 5.36 338.91 352.47
LI008 M 5.88 162.11 168.59
LIO10 M 11.17 359.75 374.14
LI003 F 11.30 353.67 367.82
LI004 F 11.93 285.55 296.97
LI006 M 12.39 489.33 508.90
LI013 F 15.67 259.72 270.11
aM, male; F, female







70


Table 4-8. Sex, age, kidney volume for the 10 children used for the kidney study.
Total Kidney Left Kidney t K Difference between Total
Age RightKidney
Case Sex (mon Volume Volume Left and Right Kidney
(ots (cm') (cm') Volme (cm) Kidney (%0o) Mass (g)
K005 F 0.03 21.46 11.67 9.79 -19 22.32
K007 M 0.23 30.65 16.32 14.33 -14 31.88
K001 F 2.17 29.86 14.89 14.97 1 31.05
K012 M 2.46 43.53 22.32 21.21 -5 45.27
K011 M 4.70 39.03 18.43 20.6 11 40.59
K009 M 5.36 36.87 18.74 18.13 -3 38.34
K008 M 5.88 39.55 20.15 19.4 -4 41.13
K003 F 11.30 89.09 46.73 42.36 -10 92.65
K004 F 11.93 50.65 25.61 25.04 -2 52.68
K006 M 12.39 95.52 52.19 43.33 -20 99.34
aM, male; F, female










Table 4-9. Mean brain weight per age period. A) For the CT-based volume data and the
Cristy and Ekerman soft tissue density (1.04 g.cm-3). B) For the ICRP
Publication 23. C) Difference between the two average masses.


Age Intervals Mean Brain Weight (g) STDa Number of cases
0 423.11 81.26 5
0-3m 469.12 129.32 22
3-6m 725.41 147.65 4
6-12m 961.11 165.01 8
1-2y 1050.65 174.57 7


Age Intervals Mean Brain Weight (g)
0 350.17
0-3m 423.83
3-6m 566.89
6-12m 803.05
1-2y 935.48


STDa Number of Cases
4.24 484
16.97 683
46.67 683
106.77 900
54.45 1061


Age Ratio CT/ICRP Difference
Ratio CT/ICRP
Intervals (%)
0 1.21 20.8
0-3m 1.11 10.7
3-6m 1.28 28.0
6-12m 1.20 19.7
1-2y 1.12 12.3


a STD, standard deviation







72


Table 4-10. Mean skull weight per age period. This was calculated from the CT-based
volume data and the linear interpolation of the Cristy and Ekerman bone tissue
densities (1.2 g.cm-3 for a newborn and 1.44 g.cm-3 for 1-year-old and above).
Age Intervals Mean Skull Weight (g) STDa Number of cases
0 128.41 35.47 4
0-3m 172.85 59.43 16
3-6m 260.22 N/A 1
6-12m 390.77 37.97 4
1-2y 460.80 38.05 4
aSTD, standard deviation











Table 4-11. Mean lung weight per age period. A) For the CT-based volume data and the
Cristy and Ekerman lung tissue density (0.296 g.cm-3). B) For the ICRP
Publication 23. C) Difference between the two average masses.


Age Intervals Mean Lung Weight (g) STDa Number of cases
0-3m 61.95 14.61 5
3-6m 97.36 36.50 5
6-9m 115.59 18.23 3
9-12m 105.61 10.14 4
1-2y 152.85 28.93 3


Number of
Age Intervals Mean Lung Weight (g) STDa e
Cases
0-3m 66.17 3.68 93
3-6m 93.70 0.57 105
6-9m 122.52 9.76 127
9-12m 142.06 0.07 112
1-2y 172.89 3.54 162


Age
Intervals
0-3m
3-6m
6-9m
9-12m
1-2v


Ratio CT/ICRP


Difference
(%)
-6
4
-6
-26
-12


aSTD, standard deviation










Table 4-12. Difference between the anthropometric models and the values determined by
the growth curves established in this study. A) Newborn stylized model. B) 1-
year-old stylized model. C) UF-newborn tomographic model. D) UF-2-month
tomographic model. E) 6-month comparison. F) 9-month-tomographic model.
A)


Organ a

C&E Brain
Bouchet's Brain
C&E Skull
Bouchet's Skull
Left Lung
Right Lung
TLV


Newborn Stylized Model
Organ Volume (cm3)
338.00
371.90
55.93
83.50
79.10
91.90
171.00


Volume determined from the
General Equationb (cm3)
374.04
374.04
87.56
87.56
67.91
99.05
167.42


I Average Differenced (%) 2.45


1-year-old Stylized Model Volume determined from the Difference0
Organ Organ Volume (cm3) General Equationb (cm3) (%)
C&E Brain 850.00 1005.70 15
Bouchet's Brain 838.40 1005.70 17
C&E Skull 161.80 282.46 43
Bouchet's Skull 252.80 282.46 11
Left Lung 225.00 199.63 -13
Right Lung 259.00 231.12 -12
TLV 484.00 430.85 -12
I Average Differenced (%) 13.16


UF-newborn
Organ Tomographic Model
Organ Volume (cm3
Brain 309.86
Skull 97.34
Left Lung 56.50
Right Lung 66.33
TLV 122.83


Volume determined from Dif Volume determined Di
b Difference Difference
the General Equation 0 from the Specific 0
) (cm3) /) Equation (cm3)e (/)
398.43 22 325.74 5
87.79 -11
78.31 28
105.43 37
183.69 33
Average Differenced(%) 22.08


Difference0
(%)
10
1
36
5
-16
7
-2










Table 4-12. Continued.
D)
UF-2-month Volume determined from Volume determined Dif
Organ Tomographic Model the General Equation D from the Specific
Organ Volume (cm3) (cm3) (%) Equatione (cm3) ( )
Brain 535.97 577.06 7 573.28 7
Skull 166.98 135.62 -23
Left Lung 74.53 111.28 33
Right Lung 80.00 134.12 40
TLV 154.53 245.32 37
SAverage Differenced (%) 22.42
E)
6-month Volume determined from Volume determined
a b Difference Difference
Organ Comparison Organ the General Equation D from the Specific Dif
Volume (cm3) (cm3) (%) Equatione (cm3) (
Brain 535.97 818.25 34 821.79 35
Skull 166.98 211.64 21
Left Lung 74.53 153.92 52
Right Lung 80.00 178.13 55
TLV 154.53 331.84 53
SAverage Differenced (%) 36.34
F)
UF-9-month Volume determined from Volume determined Di nc
Organ Tomographic Model the General Equation Di e from the Specific
Organ Volume (cm3) (cm3) E( ) Equatione (cm3) ( )
Brain 948.00 927.80 -2 950.45 0.3
Skull 323.00 251.59 -28
Left Lung 152.00 178.11 15
Right Lung 183.00 205.80 11
TLV 335.00 384.04 13
SAverage Differenced (%) 14.44
a C&E, Cristy and Ekerman's head model (Cristy 1980; Cristy and Eckerman 1987); the lung model is the
Cristy and Ekerman 's model as well.
b the general equation for the brain is the equation (4-3), for the skull the equation (4-6), for the TLV the
equation (4-7), for the left lung the equation (4-8), for the right lung the equation (4-9).
c relative organ volume differences between the model volume and the growth curve calculation.
d model average differences are calculated as indicated in Eq. (4-10).
e the specific equation for the brain is (4-4) for C), (4-5) for D), E) and F).










1400



1200



1000


0 2 4 6 8 10 12 14 16 18

Age (rnorntlis


CT-based Brain Volume
- CT-based Brain Volume Best-fit Curve


Figure 4-1. Evolution of the brain volume as a function of age over the first year and a
half of life.











1400


0
1200



1000 0
A
*
800 *



600



400



200
200 ,I --, I ,--,-i -
0 2 4 6 8 10 12 14 16 18

Age monthsl

CT-based Brain Volume
A MI RI-based Brain Volume
S CT-based Brain Volume Best-fit Curve
NMRI-based Brain Volume Best-tit Curve



Figure 4-2. Brain volume comparison between CT and MRI volumetry technique.
Evolution of the brain volume as a function of age over the first year and a
half of life for the two techniques.










1400



1200



1000


0 2 4 6 8 10 12 14 16 18


Age (months)


Figure 4-3. Evolution of the brain volume as a function of age over the first year and a
half of life for male and female.


A CT-based Female Brain Volume
* CT-based Male Brain Volume
CT-based Female Brain Volume Best-Fit Cur e
S CT-based Male Brain Volume Best-Fit Curve

































0 2 4 6 8 10 12 14 16 18


Age (lnolltlihs

CT-based Skull Volume
- CT-based Skull Volume Best-fit Curve


Figure 4-4. Evolution of the
half of life.


skull volume as a function of age over the first year and a


















































Figure 4-5. 3D reconstruction showing the evolution of the skull volume as a function of
age over the first year and a half of life. A) Skull of a newborn with the large
fontanels of connective tissues. B) Skull of a 6.31-month-old child, where all
the fontanels are still observable, but have been filled by bones except the
anterior one. C) Skull of a 13.63-month-old child. The anterior fontanel has
almost disappeared. The skull is less circular.


:; III
oil
vil
IIIIWVII'wl

0,91,11,11
wil't IN
Imll'i
It
'."11111101
IN,
IN


14 7







81


700



600 -



500 -



400 -*%

> 00
300



200 .



100 -ii
0 2 4 6 8 10 12 14 16 18

Age (ronthsli


CT-based Total Lung Volume
CT-based Total Lung Volume Best-fit Curve




Figure 4-6. Evolution of the total lung volume as a function of age over the first year and
a half of life.







































0 2 4 6 8 10 12 14 16 18
Age (months)

0 CT-based Left Lung Volume
A CT-based Right Lung Volume
CT-based Right Lung Volume Best-fit Curve
CT-based Letf Lung Volume Best-fit Curve



Figure 4-7. Evolution of the left and right lung volume as a function of age over the first
year and a half of life.














-3





-o
5 30-0



C d 0
S20-




I 10- 0*
S S




0 -
10







^I I I I I I I I
0



0 2 4 6 8 10 12 14 16 18

Age (months)

Relative Difference between the Left and the Right Lung Volume
Upper Diference between the Left and Right Lung Volume (37%)
Mean Difference between the Left and the Right Lung Volume (16"%)
Lower Difference between the Left and Right Lung Volume (-3%)



Figure 4-8. Relative difference between left and right lung volume as a function of age
over the first year and a half of life. The right lung is compared to the left
lung.







84




350

Left Lung
300
Right Lung

2- 250
E
; 200 -







50

0



Case#


Figure 4-9. Difference between the left and right lung volume for 20 children in their first

year and a half of life.












60


50 -i Ij Kii.".

U J,. -I.!'l K dq."ii.
40
E

E
30


2 20


10
0




K001 K003 K004 K005 K006 K007 K008 K009 K011 K012
Case #



Figure 4-10. Difference between the left and right kidney volume difference for 10
children in their first year and a half of life.







86






1200.00

o Brain CT 1050.65

1000.00 935.48
9 6 1 1 1 --
Brain ICRP 23

803.05
800.00
725.41



600.00

I-
469.12
423.11 423.83
400.00
350.17



200.00




0.00
0 0-3m 3-6m 6-12m 1-2y
Age Intervals



Figure 4-11. Brain mass average evolution for CT-based data and for ICRP data per age
period over the first year and a half of life.














172.89


O Lung CT


152.85
T


* Lung ICRP 23


200.00

180.00

160.00

140.00

120.00

100.00

80.00

60.00 -

40.00 -

20.00 -

0.00 -


142.06


3-6m 6-9m 9-12m
Age Intervals


1-2y


Figure 4-12. Lung mass average evolution for CT-based data and ICRP data per age
period over the first year and a half of life.


61 '







0-3m
0-3m














CHAPTER 5
CONCLUSION AND FUTURE WORK

A brief summary of the establishment of the organ-specific growth curves will be

presented hereafter followed by the conclusions that can be stated from this study.

Finally, some suggestions for the future establishment of new organ-specific growth

curves will be suggested.

Conclusions

Over one hundred CT images were successfully segmented for this study and the

organ volumes of the brain, the skull, the lungs, the liver, and the kidneys were

determined. Some errors, which are essentially due to the image quality and the

appreciation of the observer, can exist in this organ volume determination. Image

artifacts can be numerous especially with living patients and can cause problems with

semiautomatic and automatic segmentation. The presence of tubes or surgery staples

inside the patient as well as the equipment surrounding the patient that cannot always be

removed for the side of the patient can make the segmentation more difficult. The

manual segmentation helped to correct these errors.

Organ-specific growth curves were established for the normal development of the

brain, the skull, and the lungs. For symmetrical organs the differentiation between the

left and right organs (lungs and kidneys) was quantified. Furthermore, gender

differentiation was taken into account when sufficient cases were available. The graphs

of the organ evolution over time demonstrate a non-linear growth of the organ studied.

The great variation that can exist between individuals at the same age was largely









observed. Furthermore, the organs do not grow under the same schema. The brain and

the skull seem to have a very rapid growth in the first few months of life and a slower one

afterwards. The lungs, on the other hand, seem to have a relatively constant growth over

the first year of life. However, this curve could have been more accurate with a lot more

cases even though the results presented here give a good general trend of organ growth.

Furthermore, with more cases the deviation from the average could be quantified over the

time.

When comparisons are made to reference weight growth curves that were derived

primarily from autopsy studies, differences are observed. These differences can be

attributed to a combination of factors such as (1) overall increases in individual size and

stature over the past several decades, (2) problems with tissue swelling or shrinkage post

mortem, and (3) differences in techniques used for organ volume and/or mass

determination.

The characterization of the existing anthropometric models used for radiation

dosimetry purposes was done based on the growth curves of normal organ development

established in this study. The results confirmed that the stylized models were close to

representing a standard population over a large period of time, but not a specific

individual at a specific age. On another hand, the tomographic models were shown to be

a lot more specific to the individual it represents and very different from a standard

population.

Future Works

This study should be continued to fulfill a complete characterization of organ

volume growth curves. Furthermore, some improvements can be suggested to make this

occur. Increasing the number of cases studied would allow the establishment of a