Pharmacokinetics of Molecular Transport across the Blood-Brain Barrier in Pontine Gliomas

Material Information

Pharmacokinetics of Molecular Transport across the Blood-Brain Barrier in Pontine Gliomas
Brown, Alec C
Place of Publication:
[Gainesville, Fla.]
University of Florida
Publication Date:
Physical Description:
1 online resource (51 p.)

Thesis/Dissertation Information

Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Biochemistry and Molecular Biology
Committee Chair:
Mareci,Thomas H
Committee Co-Chair:
Long,Joanna R
Committee Members:
Merritt,Matthew E


Subjects / Keywords:
exchange -- glioma -- mri -- pharmacokinetics
Biochemistry and Molecular Biology -- Dissertations, Academic -- UF
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Biochemistry and Molecular Biology thesis, M.S.


Understanding the structure and function of the brain vasculature is vital to investigating neurodegenerative diseases such as Alzheimer's, or cancers such as pontine gliomas. Moreover, accurately assessing the prognosis of pontine gliomas is crucial for physicians to be able to assess treatment options. Tumors in the brain disrupt the blood-brain barrier and form immature, "leaky" blood vessels during tumor proliferation. This leakiness can be quantitatively assessed using the forward volumetric transfer constant (Ktrans) of contrast agent from blood vessels to brain parenchyma after acquiring images using dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI). In this study, rats were injected with 9L gliosarcoma cells in vivo and then tumor structure was compared across DCE, T2-weighted, T1, and fractional anisotropy (FA) images acquired in vivo at three- and ten-days post-implantation on a 4.7T magnet. Ktrans was then estimated from the DCE images and compared to their respective tumor volumes. DCE estimated the largest tumor volume at days three and ten, and tumor growth factor was correlated to the day three Ktrans constant, which consistently fell off at day ten. These results suggest Ktrans is potentially a biomarker for tumor aggressiveness which can be assessed non-invasively and could be used to develop and optimize chemotherapeutic treatments to brain tumors. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (M.S.)--University of Florida, 2019.
Adviser: Mareci,Thomas H.
Co-adviser: Long,Joanna R.
Statement of Responsibility:
by Alec C Brown.

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
LD1780 2019 ( lcc )

UFDC Membership

University of Florida Theses & Dissertations


This item has the following downloads:

Full Text




2019 Alec Brown


3 ACKNOWLEDGMENTS I first thank Dr. Mareci for providing invaluable guidance throughout my four years in the lab. I also thank Dr. Magdoom, who was a previous graduate student in the Mareci lab who collected the images used in this study prior to when I joined the lab, as well as developed in hous e software in IDL for the model and concentration maps. I also thank Dr. Rintaro Hashizume, Northwestern University, who provided the 9L gliosarcoma cells. I also thank Manish Amin, another graduate student of the Mareci lab who helped review my manuscript s and provide me guidance in my graduate years after Dr. Magdoom graduated. I thank Guita Banan and Daniel DeYoung, graduate students in the Mareci lab who set fantastic examples as graduate students throughout my years in the lab. The imaging done in this study was performed in the Advanced MRI/S (AMRIS) Facility at the McKnight Brain Institute of the University of Florida, which is part of the National High Magnetic Field Laboratory.


4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 3 LIST OF TABLES ................................ ................................ ................................ ............ 5 LIST OF FIGURES ................................ ................................ ................................ .......... 6 LIST OF ABBREVIATI ONS ................................ ................................ ............................. 7 ABSTRACT ................................ ................................ ................................ ..................... 8 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 10 2 METHODS ................................ ................................ ................................ .............. 15 Imaging Procedure ................................ ................................ ................................ 15 Contrast Agent ................................ ................................ ................................ ........ 17 Exchange ................................ ................................ ................................ ................ 19 Model Selection ................................ ................................ ................................ ...... 21 Tumor Segmentation ................................ ................................ .............................. 23 Statistical Analysis ................................ ................................ ................................ .. 23 3 RESULTS ................................ ................................ ................................ ............... 28 Imaging Comparison ................................ ................................ ............................... 28 MRI Parameter Analysis ................................ ................................ ......................... 29 4 DISCUSSION ................................ ................................ ................................ ......... 34 Defining Tumor Volume ................................ ................................ .......................... 34 K trans and Tumor Volume ................................ ................................ ......................... 35 Other Work ................................ ................................ ................................ ............. 37 LIST OF REFERENCES ................................ ................................ ............................... 46 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 51


5 LIST OF TABLES Table page 3 1 MRI parameter analysis.. ................................ ................................ .................... 32


6 LIST OF FIGURES Figure page 2 1 Signal enhancement curves.. ................................ ................................ ............. 25 2 2 Venous output function.. ................................ ................................ ..................... 25 2 3 Model of contrast agent exchange.. ................................ ................................ .... 26 2 4 Logan plot.. ................................ ................................ ................................ ......... 26 2 5 Tumor segmen tation ................................ ................................ ........................... 27 3 1 Average tumor size relative to DCE. ................................ ................................ ... 31 3 2 Model analysis example. ................................ ................................ .................... 32 3 3 Tu mor size and K trans ................................ ................................ .......................... 33 3 4 Tumor growth factor and K trans ................................ ................................ ........... 33 4 1 Correlation of micro vessel density (MVD) to K trans ................................ ............. 41 4 2 Dynamic K trans study in VX2 tumor line rabbits. ................................ .................. 42 4 3 K trans and tumor clinical stage in human prostate cancer. ................................ ... 42 4 4 Brightfield optical images of extracted brains ................................ ..................... 43 4 5 Laser confocal images. ................................ ................................ ....................... 44 4 6 Example of reconstructed perivascular network. ................................ ................ 45 4 7 Further reconstruction and extrapolation ................................ ............................ 45


7 LIST OF ABBREV IATIONS CSF Cerebrospinal fluid DCE Dynamic contrast enhanced EES Extracellular extravascular space FA Fractional anisotropy MRI Magnetic resonance imaging MVD Micro vessel density PVS Perivascular Space ROI Region of interest


8 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 PHARMACOKINETICS OF MOLECULAR TRANSPORT ACROSS THE BLOOD BRAIN BARRIER IN PONTI NE GLIOMAS By Alec Brown August 2019 Chair: Thomas Mareci Major: Biochemistry and Molecular Biology Understanding the structure and function of the brain vasculature is vital to cancers such as pontine gliomas. Moreover, accurately assessing the prognosis of pontine gliomas is crucial for physicians to be able to assess treatment options. Tumors in the brain disrupt the blood during tumor proliferation. This leakiness can be quantitatively assessed using the forward volumetric transfer constant (K trans ) of contrast agent from blood vessels to brain parenchyma after acquiring images using dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI). In this study, rats were injected with 9L gliosarcoma cells in vivo and then tumor structure was compared across DCE, T 2 weighted, T 1 and fractional anisotropy (FA) images acquired in vivo at three and ten days post implantation on a 4.7T magnet. K trans was then estimated from the DCE images and compared to their respective tumor volumes. DCE estimated the largest tumor volume at days three and ten, and tumor growth factor was correlated to the day three K trans constant, which co nsistently fell off at day ten. These results suggest K trans is potentially a biomarker for tumor aggressiveness


9 which can be assessed non invasively and could be used to develop and optimize chemotherapeutic treatments to brain tumors.


10 CHAPTER 1 INTRODUCTION Magnetic resonance imaging (MRI) is an invaluable tool to investigate tumors as this imaging method allows robust, non invasive imaging of tumors. Biological models of interest are primarily of water, which con tains two protons per molecule. When biological subjects such as rats or humans are in a strong magnetic field, such as a magnetic field generated by a 3 Tesla magnet, the proton nuclear spins will align with the magnetic field. To produce images, once pro tons are aligned with a magnetic field, radio waves are pulsed at a frequency dependent on the strength of the magnetic field realign with the local magnetic field produced by the radiofrequency coils. When these coils are turned off, the protons then realign with the static magnetic field. The relaxation rates of protons returning to their resting states can be measured using receiver coils which measure the radio waves emit ted by protons. Two types of relaxation can be examined as two the relaxation rate of the bulk magnetization vector back to equilibrium (T1 relaxation) and the loss of coherence of the proton spins (T2 relaxation). Differences in the local magnetic enviro nments of protons due to their surroundings, such as local fat cells will cause the relaxation times of the protons to differ and allows the creation of an image of the biological subject when deciphering the resulting radiofrequency waves [1] The use of contrast agents such as gadolinium labeled albumin (GD ab), which are injected locally to the area of interest, enhance the measured signal due to changing the magnetic environment of the local protons and ca using a change in the relaxation rates. The resulting increase in relaxation rates


11 causes a measurable enhancement in the produced signal, which allows the study of dynamics of tumor growth as well as tumor volume. In healthy brain tissue, blood vessels ar e ensheathed by the blood brain barrier, which is a border between circulating blood and the extracellular extravascular space formed by tight junctions between endothelial cells [2] The endothelial cells are supported by glial cells such as astrocyte end feet which surrou nd the endothelial cells and supply these cells with nutrients and maintain homeostasis [3] In gliomas, the supportive glial cells grow and proliferate out of control due to causes such as radiation or hereditary disorders. This abnormal c ell growth causes a breakdown of the blood brain barrier and allows circulating molecules to leak into the brain parenchyma. Gliomas compromise 30% of brain tumors and 80% of malignant brain tumors. These tumors are highly aggressive, with a reported 5 yea r survival rate of 15 35% [4] As these tumor cells proliferate, the demand for nutrients increases to meet the rapid growth. Thus, the malignant cells releas e growth factors such as vascular endothelial growth factor (VEGF), which stimulates the formation of new blood vessels, a process referred to as angiogenesis [5] The newly formi ng vessels are generally abnormal, vessels differ structurally in that they have no pericyte and basement membrane, the vessel diameter is smaller, the vascular density is heter ogenous, and the permeability to large molecules is higher [6] Due to these differences, the neovasculature is less structurally sound and is susceptible to leaking c irculating molecules, such as contrast agents [7] Previous studies have found that assessi ng VEGF expression in a tumor,


12 and thus, angiogenesis, the aggressiveness of a tumor can be quantified [6] However, MRI represents a quicker, non invasive alternative to assessing tumor qualities. In MRI, as previously stated, contrast agents are commonly used to enhance signal localized to where the agent is delivered. In growing tumors, due to angiogenesis th e contrast agent leaks from blood vessels into the extracellular extravascular space. The accumulation of contrast agent in the extracellular extravascular space can be quantified by analyzing the local MR voxels and assessing the change in relaxation rate s caused by the local contrast agent. By normalizing the accumulated contrast agent in the voxel relative to the influx of contrast agent in the blood due to the bolus injection, the forward volumetric transfer constant ( K trans ) of the exchange of contrast agent from the vascular compartment of the voxel to the extravascular extracellular space compartment can be mathematically extracted. This transfer constant is [8] Recent studies that have used K trans to investigate tumors have found a correlation between K trans coefficien ts and tumor invasiveness. For example, a retroactive clinical study done by Yeo et al. [9] found that in rectal cancer K trans was directly correlated to the stage of invasion of the tumor. An other study done by Li et al. [10] invest igated the changes in dynamic contrast enhanced (DCE) MRI parameters such as K trans and MVD (micro vessel density, another biomarker for angiogenesis) over time in rabbits implanted with VX2 tumor cells. They found that K trans was directly correlated to MV D in growing tumors, and that K trans peaked during tumor growth and then dramatically fell off over time. However, there is a lack of understanding of the


13 relationship between K trans and tumor size, and glioma MR studies have not assessed this relationship using fixed time intervals. More specifically, in the case of gliomas, multiple studies have found K trans to be associated with the histopathological tumor grade (which is a marker for tumor aggressiveness), in that high grade gliomas were distinguishable from low grade tumors by their higher K trans [11, 12] Another study by Subashi et al. [13] which used an extended Tofts model to assess K trans in brainstem gliomas found tumor volume to be positively correlated to K trans however this study did not analyze tumor dynamics at fixed time intervals. Choi et al. [14] investigated the potential prognostic use of K trans and found that higher K trans values were associated with lower survivability. These results highlight the potential use of K trans as a biomarker for tumor aggressiveness, and further understanding of the relationship between K trans and tumor properties is vital for effective prognosis. This study sought to analyze the relevance of these MR parameters to glioma prognosi s and address the inconsistencies of the reported relationship between K trans and tumor size. Accurate tumor volume and definition is also crucial for surgical procedures, as an optimized imaging protocol for types of cancer such as gliomas would provide c linicians with information necessary for effective treatment. Wood et al. [15] studied the prognostic value of tumor volume and found that post operational tumor volume was indicative of survivability, which shows the need for optimizing imaging protocols for t umors. Beyond being useful in prognosis, K trans could also be useful for assessing the most efficient way to deliver chemotherapeutics to the tumor [16] For example, in the


14 brain, the blood brain barrier represents a significant hurdle for dru g delivery as the blood brain barrier is impermeable to most large molecules. However, in a tumor, the leakiness of newly forming vessels could possibly provide an optimal window for drug delivery, and quantifying tumor properties using K trans would be par amount to the development of effective brain tumor chemotherapeutic treatments.


15 CHAPTER 2 METHODS Imaging Procedure Following the protocol developed by Jallo et al. [17] and Barth et al. [18] 9L gliosarcoma cells were cultured in T175 cell culture flasks in Dulbe penicillin streptomycin. The cells were then incubated at 37 degrees Celsius in a 5% CO 2 atmosphere. Prior to surgery, the cells were harvested using 0.25% trypsin EDTA, washed twice, and suspended in a serum free medium until achieving a concentration of 3 X 10 4 Experiments were performed on two month old Sprague Dawley rats (n=9) weighing ~250 g using protocols and procedures approved by the Institutional Animal Care and Use Committee. The rats were first anesthetized using 10 mg/kg subcutaneously applied xylazine and 4% isoflurane in 1 L/min oxygen and maintained throughout surgery using 1.5% isoflurane in 1 L/min oxygen. Prior to imaging, 2 mL of saline was injected subcutaneously for hydration. Following anesthetization, the rats were fixed in position using a stereotaxic frame (Kopf Model 900, David Kopf Instruments, Tujunga, CA), and a burr hole was drilled above the pontine reticular formation (AP: + 1.0 mm, ML: 1.4 mm and DV: 7.0 mm from the interaural line). A 32G needle (small hub RN needle, Hamilton, Reno, NV) was syringe (Hamilton). Finally, using a syringe pump (Cole solution contai ning ~10 5 tumor cells were infused into the pontine reticular formation at


16 MR images were obtained in vivo at 3 and 10 days post implantation. To prepare the rats for imaging, the tail vein was catheterized, and a 1 mL tuberculin syringe was connected to the catheter via a ~1.5 m long intravenous line for bolus injection of the contrast agent. Rats were then fixed in a 39 mm ID quadrature transmit/receive birdcage RF coil (Animal Im aging Research, Holden, MA) and imaged on a 330 mm ID Oxford 4.7T horizontal bore magnet with a RRI BFG 200/115 S14 gradient set (Resonance Research, Billerica, MA) connected to an Agilent VNMRS console with VnmrJ3.1A software (Agilent Technologies, Santa Clara, CA). Once fixed in the bore, the syringe was placed external to the bore to allow bolus injection of 0.24 mmol/kg body weight of gadolinium based gadodiamide (Omniscan, GE Healthcare, Milwaukee, WI) by hand. Respiration and heart rate were monitored throughout imaging using a pneumatic pillow attached to an animal monitoring system (SA Instruments, Model 1025, Stony Brook, NY). T 1 weighted, T 2 weighted DCE, and FA images were then acquired using ten 1 mm axial slices with a field of view of 24 x 24 mm and a matrix size of 96 x 96. T 1 was quantified using an inversion recovery based rapid acquisition with relaxation enhancement (RARE) sequence using TR = 10 sec, TE eff = 8.51 msec RARE factor = 4, TI = 100, 500, 1000, 1500, 2000, 4000 msec, and NEX = 1. T 2 weighted images were acquired using a RARE sequence with TR = 10,000 msec, TE eff = 120 msec, RARE factor = 4, and NEX = 3. High angular resolution diffusion imaging (HARDI) data was acquired using a fat suppressed, spin echo diffusion weighted, multishot echo planar imaging (EPI) sequence using TR = 5000 msec, TE eff = 40 msec, number of s cams = 3, and NEX = 8. Diffusion weighted images were taken in 24 directions distributed usin g a


17 scheme of electrostatic repulsion [19] (six with b 2 and 18 with b value 2 14 msec. Using in house software written in IDL (Harris Geospatial Solutions, Broomfield, CO), diffusion weighted images were fitted to a rank 2 tensor model, and then scalar FA values were obtained. For DCE images, a serial T 1 weighted spoiled gradient echo sequence (TR = 100 msec, TE = 2.39 msec, flip angle= 90 degrees, NEX = 3, temporal resolution = 29 sec per slice pack) was used. Three prescans were first acquired prior to the injection of gadodiamide, and 122 scans were acquired post injec tion. Contrast Agent The contrast agent being used, gadodiamide, does not cross the blood brain barrier because of its large size and hydrophilicity [20] In healthy brain vasculature, the contrast agent flows through blood vessels in the brain without entering the extracellular extravascular space (EES). However, contrast agent will leak across the blood brain barrier into the EES in immature, unstable blood vessels formed by angiogenesis during tumor proliferation. Thus, to visualize and model this leakiness, the voxels of brain tissue were modeled as two compartments: bl ood plasma and EES. To estimate the concentration of gadodiamide in tissue voxels, the MR signal enhancement caused by the presence of gadodiamide was first converted to concentration estimates using a method described by Chen et al. [21] MR signal, S, from a conventional spoiled gradient echo sequence is given by: (2 1)


18 0 2 is the reversible transverse relaxation time constant Fast exchange of protons between blood and EES allows us to assume the tissue relaxes with a single relaxation time, and thus the effect of contrast agent on the relaxation rate of nearby nuclei can be calculated using E quation 2 2 : (2 2) Where T 10 is the T 1 relaxation time of the tissue without contrast agent, r 1 is the relaxivity of contrast agent, and C is the contrast agent concentration U sing E quation 2 1 and E quation 2 2 MR signal enhancement due to a contrast agent is as follows: (2 3) Ignoring transvascular and transcellular water exchange effects on relaxivity allows us to assume r 1 is constant. Equation 2 3 can then be rearranged to yield an estimate of contrast agent concentration in the voxel: (2 4) Relaxivity values for rats at body temperature (36 37 degrees Celsius) and at 4.7 T are 2.8 sec 1 mM 1 in brain tissue and 3.9 sec 1 mM 1 in blood plasma [22, 23] To avoid inaccuracies in the measured T 1 caused by flow, blood vessel voxels were assigned a constant T 1 0 o f 2.7 seconds [22] The T 10 for tissue was calculated by fitting the polarity cor rected signal magnitude from the variable inversion time RARE acquisition to the standard inversion recovery model. Using E quation 2 4, estimated


19 concentration maps were generated in IDL at the time of peak enhancement for each tumor. To verify relative si gnal enhancement in the tumor, region(s) of interest (ROI) were placed at the center of the tumor, the healthy contralateral side of the pons from the tumor, the superior sagittal sinus, and muscle. Signal enhancement curves were then generated for these R OIs, as shown in F igure 2 1. The contralateral ROI was used as the control for normalizing signal enhancement, as in healthy brain tissue there should be no leaking of the contrast agent into the brain parenchyma. The blood ROI peaks and then quickly decay s, which is expected as it corresponds to the bolus injection of gadodiamide and its distribution into the vasculature. The muscle ROI was shown to display signal enhancement relative to the control, which is expected as muscle does not have the blood brai n barrier and thus some exchange of contrast agent across the blood vessels will occur. The tumor ROI exhibited the largest amount of signal enhancement, displaying the expected enhanced vascular permeabi lity as a result of angiogenesis. Exchange After the concentration of contrast agent was calculated, this concentration was fit to a two compartmental exchange model developed by Tofts and Kermode [8] Leakiness of contrast agent into the EES was assumed to be small relative to the flow of contrast agent in blood plasma which allows tissue voxels to be modeled as two distinct compartments. Assuming the contrast agent is well mixed in each compartment, the exchange of contrast agent between compartments can be estimated as follows: (2 5)


20 Where C p is the concentration of contrast agent in blood plasma, C e is the concentration of contrast agent in EES, C t is the concentration of contrast agent in tissue, v e is the volume fraction of EES in the tissue, V t is the volume of tissue, P is the permeability coefficient of the blood vessel, and S is the surface area of leaking blood vessel. Using C t the concentration of contrast agent in the entire tissue voxel which is a weighted sum of concentrations in each compartment (C t = v e C e + v p C p ), E quation 2 5 can be rewritten in terms of the contrast agent concentration estimated from E quation 2 4. Substituting K trans = PS/V, E quation 2 5 further simplifies to : (2 6) Equation 2 6 can be further simplified when v p (volume fraction of blood plasma in tissue) < < v e K trans /v e can also be recognized as the exchange of contrast agent from the EES to the blood plasma compartment k ep The further simplified e quation is then: (2 7) The solution to E quation 2 7 can be obtained using the extended Tofts model [24] : (2 8) Where is the is the Dirac delta function, and I(t) is the impulse response of tissue to short bolus of contrast agent: (2 9) This model assumes that the system is linear and tissue parameters are time invariant. As rat physiology is monitored and steadily maintained throughout the experiment, the biological system can be assumed to function time invariantly. At high blood concentrations of contrast agent, the system becomes nonlinear due to the nonlinear relationship between the concentration of contrast agent and T 2 dephasing


21 [24] To ensure the system remains linear, the dose of contrast agent was selected based on the r 1 of the contrast agent and T 10 of the blood. To estimate the concentration of contrast agent in the blood plasma, a venous output function (VOF) was used. While arterial input functions (AIF) are generally standard for estimating C p [8] measuring contrast agent in fast flowing arterial blood is more erroneous as the signal loss due to T 2 dephasing is greater [25] Moreover, the shape of the AIF has been shown to match the shape of the VOF [26] To estimate the shape of the VOF for every rat, an ROI was placed in the superior sagittal sinus. To estimate the concentration of contrast agent in the blood plasma of the caudate putamen, where the tumors are proliferating, the amplitude was adjusted with a 1% plasma volume fraction [27] An example of the placed ROI and resulting VOF is shown in F igure 2 2. The VOF was approxima ted as biexponential, as shown below [8] : ] (2 10) Where a 1 is the amplitude of fast equilibrium between blood plasma and EES, m 1 is the rate constant of fast equilibrium between blood plasma and EES, a 2 is the amplitude of clearance, m 2 is the rate constant of clearance, and d is the dose of bolus injection. Model Select ion To account for the variability in tissue environment, a multimodal approach developed by Ewing et al. [24] was used. In this approach, fou r different cases were used to fit the experimental data. The model is shown in F igure 2 3. Case 1 corresponds to necrotic tissue in which there is no exchange due to vasculature degradation (K trans = K ep = v p = 0). Case 2 accounts for healthy brain tissue in which there is no exchange across the BBB and contains healthy vasculature (K trans = K ep = 0,


22 v p fit experimental data at the boundary of the tumor, in which the EES volume fraction of the voxel is small enough to appear as if there is no reflux (K trans ep = 0, v p Case 4 is the model of interest, in which the blood vessels are leaky, and reflux is borne in the data (K trans ep p e which case each voxel represents, MR measured concentration time series data from each voxel was fitted to each of the four cases using a nonlinear least squares analysis in IDL (MPFIT [28] ). Akaike information criterion (AIC) was then used to calculate the probability that a voxel fits a particular case, and then the voxel is labeled as the most probable case [29] Using the AIC weights for each voxel fitted to every case weighted parameters K trans K ep and v p are estimated. The approach for AIC is shown below: (2 11) (2 12) (2 13) Where SSR i is the sum of square residual for case i, N is sample size, k i is the number of parameters in case i, and p is the fitted parameter (K trans K ep v p ) To determine the number of fitted points, N, a Logan plot was used to graphically confirm the end timepoint in which the accumulated contrast agent concentration reached equilibrium in each compartment [30] To generate the Lo gan plot, an ROI was first placed in the center of the tumor, and E quation 2 14 was integrated across the ROI to estimate the accumulated contrast agent in the voxel relative to the accumulated


23 contrast agent in the blood plasma compartment. The nonlinear portion of the Logan plot represents pre equilibrium in the tissue and contains the fitted points of interest. An example of the tumor ROI and corresponding Logan plot is provided in F igure 2 4 (2 14) Tumor Segmentation To quantify the volume of the pontine tumors and the parameters corresponding to those tumors, the tumors were segmented from the acquired quantitative T 1 maps, T 2 weighted images, FA images, and DCE images at peak signal enhancement. The used to place a seed ROI at the center of the tumor to grow to surrounding pixels with a signal intensity within 1 to 3 standard deviations of the average signal intensity of the seed. The seed size, seed placement, and standard deviation multiplier was selected to contain the region to the pons, and to exclude cortical tumor regions and healthy tiss ue. The adherence of the region to the tumor was confirmed visually. An example of seed placement and grown ROI for each imaging technique is provided in F igure 2 5 Cortical tumors were separately segmented for further analysis. Statistical Analysis To an alyze the collected data, a Wilcoxon rank sum test was used to test for statistical significance of estimated tumor volumes and MRI parameters (K trans v e ) against noise, as well as to test the statistical significance of differences between T 1 based, T 2 based, FA based, and DCE based tumor volumes at days 3 and 10. This test was also performed to assess the statistical significance of differences in MRI parameters. The Wilcoxon rank sum test and Wilcoxon signed rank test were chosen


24 due to the small sampl e size (n=9) used in this experiment, as well as because these tests do not assume a normal distribution of the experimental data [31]


25 Figure 2 1. Signal enhancement curves. A) Regions of interest placed in the superior sagittal sinus (blood), the tumor, healthy contralateral side (control), and muscle. B) Signal enhancement curves calc ulated relative to contralateral region. This image was generated by Dr. Magdoom. Figure 2 2. Venous output function. A) Example of region of interest placed in the superior sagittal sinus to estimate venous output function. B) Venous output function curve following bolus injection of gadodiamiade.


26 Figure 2 3. Model of contrast agent exchange. Voxels are modeled as two compartments and exchange between these two compartments is modeled as a solvable differential equation. Figure 2 4. Logan plot. A) Region of interest placed in center of tumor to calculate Logan plot. B) Generated Logan plot. The linear portion is when the contrast agent is in equilibrium between the extracellular extravascular space and blood plasma.


27 Fi gure 2 5. Tumor segmentation. The top row of images is the same tumor with seed placement across every imaging method, and the bottom row is the resulting


28 CHAPTER 3 RESULTS Imaging Comparison The results of this study were published in the Journal of Magnetic Resonance Imaging [32] A bar graph comparing the estimated relative average tumor volume for each imaging technique is provided in F igure 3 1. While tumors were visible at day 10 for every imaging method, day 3 t umors were primarily visible in DCE images, and only DCE imaged tumor volumes were statistically significant when compared to noise (p=0.009). At day 3, the tumors were generally too small to be discernable from noise, and even at low standard deviation mu ltipliers, containing the tumor ROI to exclusively the tumor was not possible except for larger day 3 tumors using DCE images. The contrast enhancement provided by gadodiamide proved to be essential in defining the tumor boundary using the region grow func tion. At day 10, all the employed imaging techniques produced statistically significant results when compared to noise (p=0.0039 for T 2 T 1 DCE, p=0.0091 for FA) When compared across imaging techniques, the tumor volumes estimated by T 1 T 2 and DCE image s were all shown to be significantly different to the tumor volume estimated from FA (p=0.031 for T 1 p=0.011 for T 2 and p=0.0035 DCE). There were no significant differences in the estimated average volume between T 1 T 2 and DCE images. As shown in F igure 3 1, the FA images consistently estimated a smaller tumor volume compared to other imaging methods. As with day 3 tumors, the tumor boundary was not well defined enough in FA to be discernable from healthy tissue in most cases, even at small standard devi ation multipliers. The smaller estimated tumor volume as well as the indiscernible tumor boundary can be explained by tumorigenesis producing abnormal


29 diffusion at the tumor boundary due to the disrupted vasculature and white matter tracks. MRI Parameter A nalysis Examples of the fitted model map as well as EES volume fraction and K trans maps are provided in F igure 3 2 for 3 example tumors small (2.12 mm 3 day 3), medium (11.31 mm 3 day 10), and large (19.87 mm 3 day 10). As shown, cortical tumors were typi cal in our rats, particularly for rats with large pontine tumors. Cortical tumors are expected, as the implantation of the tumor cells penetrates and disrupts the tissue, allowing tumor cells to seed and proliferate along the needle tract. The cortical tum ors were not included in the tumor volume analysis due to large differences in microstructure and vasculature [33] but the cortical tumors were included in the parameter comparison. However, cortical tumors did not visually appear at day 3 and were only visible in 6 rats at day 10. As shown, cases 3 and 4 dominate the tumor in the model map case 3, which is the exchange model without reflux, at the tumor boundary and case 4, which is the full exchange model, in the center of the tumor. As previously stated, case 3 is expected to appear at the tumor boundary, as blood vessel leakage is not actually occurring without reflux but is instead due to partial volume effects. Thus, case 3 was excluded from the analysis and instead our analysis focused on exclusively case 4. The results for average K trans and EES volume fraction at da ys 3 and 10, as well as cortical tumors at day 10, are shown in T able 3 1. The average K trans and EES volume fraction at days 10 and 3 were found to be statistically significant when compared to noise (p=0.0039 for both parameters, across both days). As sh own, the average K trans was reduced at day 10 by ~41% relative to day 3, from 0.034 min 1 to 0.02 min 1 (significantly different, p=0.08). The average EES volume fraction was also


30 12% larger at day 10, from 0.044 to 0.039 (not significantly different, p=0. 6). The cortical tumors exhibited a K trans of 0.10 min 1 which was found to be statistically significant compared to noise (p=0.031) and significant compared to the day 10 K trans for the pontine tumors (p=0.0004). The average EES v f was also significant compared to noise (p=0.031) and was significant compared to the average EES v f of pontine tumors at day 10 (p=0.0048). The relationship between tumor volume and K trans was also ana lyzed. Figure 3 3 shows tumor volume plotted against average K trans for all segmented tumors (n=18) fit to a trendline that corresponded to the highest correlation value. Average tumor K trans was shown to decrease with tumor volume (R 2 =0.34, p=0.011) and f it to a power trendline. To further assess the relationship between K trans and tumor volume, K trans at day 3 was plotted against the tumor growth factor (tumor volume at day 10 divided by tumor volume at day 3 and fit to the trendline that produced the hig hest correlation value, as shown in F igure 3 4. K trans at day 3 was found to be positively correlated to tumor growth factor (R 2 =0.46, p=0.045), and fit to a linear trendline.


31 Figure 3 1. Average tumor size relative to DCE. Tumor volumes compared across imaging methods and days. The error bars represent the 95% confidence interval. The statistical significance compared to noise was calculated using the Wilcoxon signed rank test, and the st atistical significance between samples was calculated using the Wilcoxon ranked sum test.


32 Figure 3 2. Model analysis example. A small (2.12 mm 3 day 3), medium (11.31 mm 3 day 10), and large (19.87 mm 3 day 10) sized tumor were selected to present th e corresponding maps generated by the model analysis. First, voxels are categorized into cases: case 1 necrotic tissue, in dark blue; case 2 healthy tissue, in light blue; case 3 tumor boundary, yellow; case 4 vascularized, leaking tissue with reflux, red. The following row is a map of the calculated K trans values, followed by a map of the extracellular extravascular space volume fraction. This image is F igure 5 from Magdoom et al. [32] Table 3 1. MRI parameter analysis. All parameters were statistically significant compared to noise. The average K trans was found to be significantly different between days 3 and 10, with a decrease of 41%. The difference between the day 3 and day 10 average EES v f was not statistically significant. The difference between cortical and pontine average K trans a verage EES v f was statistically significant. Average K trans (min 1 ) Average K trans standard deviation (min 1 ) Average EES v f Average EES v f standard deviation Day 3 0.020 0.016 0.039 0.015 Day 10 0.034 0.0075 0.044 0.016 Day 10 cortical 0.10 0.091 0.086 0.062


33 Figure 3 3. Tumor size and K trans K trans was shown to decrease with tumor size. The correlation coefficient was found to be statistically significant (p=0.011). The error bars represent the 95% confidence interval for K trans values. Figure 3 4. Tumor growth factor and K trans Higher K trans values at day 3 were found to be correlated to a larger change in tumor volume at day 10. The correlation coefficient was found to be statistically significant (p=0.046). The error bars represent the 95% confidence interval for K trans values. R = 0.34 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 5 10 15 20 25 30 Ktrans (min 1 ) Tumor size (mm 3 ) Tumor size vs Ktrans Day 3 Day 10 R = 0.4585 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 5 10 15 20 25 30 35 K trans at day 3 Volume of tumor at day 10 / Volume at day 3 Day 3 K trans vs tumor growth


34 CHAPTER 4 DISCUSSION Defining Tumor Volume Accurately imaging tumors is a vital process in prognosis and treatment. A study by Wood et al. [15] in which pre and post operative scans of 510 patients with malignant gliomas were studied in relation to patient survivability. Wood et al. found that post operative tumor volumes were strongly correlated to survivability (P < 0.0001). These results highli ght the importance of optimizing the definition of the tumor boundary, as a more successful resection (in that the post operational tumor volume is minimized) is correlated to survivability. Our study found that dynamic contrast enhanced images of gliomas, when compared to T 1 T 2 and FA images, estimated the largest average tumor volume at day 3, suggesting that DCE imaging is best equipped as a prognostic tool for tumor resection surgeries in earlier tumor stages. At day 10, the imaging methods did not di ffer significantly beyond FA, which estimated a significantly smaller average tumor volume. This may be due to the effects of tumorigenesis at the boundary of the tumor, as changes within the vasculature and microstructure of tissue within voxels at the tu mor boundary may presuppose significant changes in FA, allowing proper visualization of the tumor boundary. To more fully optimize imaging techniques for estimating tumor volume, histological analysis of the tumor volume using optical imaging techniques s uch as bioluminescence imaging or fluorescence imaging is required to cross check the accuracy of estimated tumor volume by imaging technique. While the largest estimate of tumor volume would prove to be the most successful in tumor resections, inclusions of areas at the tumor boundary in the case of imaging techniques such as DCE imaging


35 could lead to over estimating tumor volume and thus introduce further complications in patients by removing critical brain tissue that did not need to be removed. K trans and Tumor Volume Our study found K trans to decrease with tumor volume, contradicting a comparable study done by Subashi et al. [13] s model to assess K trans in gliomas. Another finding of our study was that K trans significantly fell off from day 3 to day 10, suggesting that K trans could correlate to tumor stage as opposed to tumor volume, which would explain the contradiction of our re [13] which also modeled tumor dynamics using the Tofts exchange model, as in that study rats were imaged following the onset of glioma sym ptoms (lethargy, head tilt, increased head size) as opposed to at a fixed time. A study on rabbit VX2 tumors by Li et al. [10] reported micro vessel density (MVD, a biomarker for angiogenesis), K trans and tumor volume over the course of 4 weeks. K trans was found to positively correlate with MVD, as shown in F igure 4 1. This is an expected result, as a larger density of leaky vessels should produce a higher K trans for a given voxel. But the interesting finding of this study is that K trans was found to increase with time, and then slowly decrease with the onset of necrosis, as shown in F igure 4 2. These findings, while in a different tumor model and animal, could explain the trends both Subashi et al. and our study found. K trans would be ex pected to be high as the tumor rapidly proliferates but tapers off upon vessel maturation and vascular collapse with the onset of necrosis. Our study did not visualize the onset of necrosis, as no significant volume of any tumor at day 10 was sorted into m odel case 1. However, the results we obtained could be explained by the onset of necrosis, as vascular collapse would predate necrosis, and would result in lower K trans values in maturing tumors. Our study found that cortical tumors exhibited a significant ly


36 higher average K trans value, which is expected, as the cortex is more highly vascularized and thus can support a larger amount of contrast agent leakage into brain parenchyma. ed cortical gliomas on average had larger K trans values as well as larger tumor volumes in comparison to pontine gliomas. Multiple previous studies have found that K trans was higher in higher histological grades of gliomas, allowing glioma grades to be dif ferentiated by K trans [11, 12] As the histologic al grade for gliomas is indicative of tumor aggressiveness [33] K tran s was found to be a biomarker for tumor aggressiveness. Our study corroborates this, as we found that the day 3 K trans was positively correlated to tumor growth factor, also confirming the utility of using K trans as a prognostic tool. It should be noted, h owever, that this data did not have a strong correlation, possibly due to the small sample size (n=9). Another study by Wu et al. [34] found the clinical stage of prostate cancer in humans to be correlated to K trans Figure 4 3 shows the K trans values sorted into their corre sponding tumor stage. Again, these results could possibly be explained by the differences in tumor line as well as animal model, but when taken into consideration relative to the results of experiments performed on gliomas, these results seem to corroborat e the model being put forward that K trans is a biomarker for tumor stage and aggressiveness. Moreover, K trans has been found to be an idiopathic parameter across different types of tumors a study by Little et al. [35] found that K trans did not vary significantly across different cancers except for gliomas. This study found that K trans was an idiopathic biomarker for vascular permeability, thus explaining the difference for gliomas, as the blood brain barrier represents a significan t barrier to vascular


37 permeability. Therefore, studies of different cancer lines and their respective K trans are comparable, as the underlying pathology remains similar, but should be done cautiously in the case of gliomas. One of the limitations of this study was the amount time points sampled. As frequent tail injections and imaging procedures decrease the survivability of the rat [36] only two time points were used. As a result, necrosis at the tumor core wa s not visualized and the timeline for changes in K trans and tumor volume was not fully established. Further studies should investigate the relationship between time from tumor implantation and K trans as previous studies in conjunction with this study seem to indicate a model in which K trans is related to tumor stage as opposed to tumor volume. Early tumor stage K trans should also be further investigated, as the results of this study suggest the high K trans present at day 3 is due to higher rates of angioge nesis, thus providing the structure and support for higher rates of cell proliferation. The higher K trans at day 3 is also indicative of an optimal window for drug delivery the ability of chemotherapeutic drugs to leak into the brain parenchyma falls off over time and thus should be administered at earlier tumor stages, although further investigation is needed to find the optimal drug delivery window, which should correspond to the tumor stage with peak K trans values. Other Work A significant portion of my time in the lab was dedicated to another project with Dr. Magdoom, which was accepted to be published by Scientific Reports in 2019. The project and its results are briefly summarized below. Understanding metabolic waste clearance in the brain is vital to investigating


38 studies have hypothesized that waste clearance in the brain is driven by cerebrospinal fluid (CSF) exchange with interstitial fluid in the brain parenchyma [37, 38] CSF is supplied from annular spaces called perivascular spaces (PVS) surrounding cerebral blood vessel s penetrating from the subarachnoid space [37] This exchange of CSF in the brai n parenchyma is indicative of a pathway that excludes the blood brain barrier, thus investigating perivascular spaces could potentially provide a method of drug delivery to the brain parenchyma. Furthermore, recent studies have found that sleep enhances me tabolic waste clearance in the brain, explaining the physiological need for sleep [39, 40] To further understand the role of sleep as well as neurode generative diseases, a method to visualize perivascular spaces in the brain is necessary. In this study, a method was developed to visualize the perivascular network in the whole rat brain. Magnetic resonance (MR) visible tracer (Gadolinium diethylene tri amine pentaacetic labeled human serum albumin) was infused in vivo into the CSF filled lateral ventricle in rats (n=5), followed by ex vivo MR imaging of the extracted brain wrapped in gauze at 17.6 T using a 40 m isotropic T 1 weighted 3D spoiled and phase re wound gradient echo dual echo imaging sequence (TR = 100 ms, TE = 3/15 ms, flip angle = 50 o 20 x 16 x 12 mm 3 FOV, matrix size = 500 x 400 x 300, 7 averages). The acquired images were then segmented to reconstruct the whole brain perivascular network. The extracted brains were then preserved in a solution of 30% sucrose in 0.1M PBS. Brightfield optical images of the extracted rat brains a re shown in F igure 4 4, imaged using dissection microscope eyepiece camera. As this signal enhancement could also potentially result from the tracer entering the bloodstream and binding to the


39 luminal wall of blood vessels, laser confocal images of a hepa tic blood vessel from a tracer infused rat were acquired using a Nikon A1RMPsi STORM 4.0 system (Nikon Instruments, Inc, Melville, NY, USA). If the tracer was binding to the luminal wall of blood vessels, the tracer would be expected to be present in blood vessels within regions other than the brain, such as the liver. As shown in Figure 4 5, confocal images show that the tracer was visualized in perivascular spaces throughout the brain. In Figure 4 5, Part C, an image of a hepatic blood vessel lacks fluore scence resulting from the presence of the tracer, confirming that the tracer did not enter the bloodstream and bind to the luminal wall of blood vessels. Following image acquisition, the images were then processed to reconstruct the perivascular network in 3D. Areas of the brain that contained the contrast agent experienced T 1 shortening due to the gadolinium tracer. As a result, perivascular spaces that the tracer was localized to appeared bright relative to the rest of the image. Moreover, as a result of fixation, the rat brain vasculature appeared dark relative to the rest of the image. To visualize the perivascular network, the rat brains were first segmented using rodent brain extraction tool (rBET) [41] in FSL software [42] and ITK SN AP software [43] to extract the cerebrum as well as remove image artifact s due to the gauze surrounding the brain during imaging. The resulting rat brains were then [44] using FMRIB's linear image registration tool (FLIRT) [45, 46] in FSL software to assess perivascular space anatomy. Maximum intensity projections of the resulting images were then acquired to allow a full visualization of the rat brain perivascular network.


40 The rat brain vasculature was further visualized by inverting the acquired T 1 weighted image stacks in FIJI [47] Inverting the perivascular images caused the dark regions in the image (blood vessels and air pockets) to appear bright and have a non zero intensity value To segment out the blood vessels from the rest of the image stack, each point in the image is by convolving the image stack with a second derivative Gaussian function to create a Hessian matrix that describes image curvature [48] This plug in is commonly used to visualize blood vessels in 3D image stacks. The original stacks of perivascular images were then thresholded by 99% to visualize areas of increased signal enhancement and thus the PVS network. Overlaying the thresholded perivascular images onto the vasculature achieved from tubeness provided a reconstruction of the rat brain perivascular network relative to the vasculature. Examples of th e reconstructed vasculature and overlaid perivascular spaces are shown in F igure 4 6. A further extrapolated reconstructed image of the vessel shown in F igure 4 6 is provided in F igure 4 7. This study was successful in reconstructing the whole rat brain perivascular network. This study has developed a tool to further investigate the perivascular network by enabling the study of glymphatic pathways as well as perivascular connectivity througho ut the brain. This high resolution MRI methodology achieved a voxel size two magnitudes smaller than previously reported [39] Further studies should investigate the comparing structural diffe rences with healthy brains. Studying structural changes in the perivascular network during sleep could also more clearly reveal the physiological role


41 of sleep. Moreover, as the blood brain barrier currently impedes drug delivery to the brain parenchyma, m odeling drug delivery using the perivascular network potentially provides an alternative approach to successfully delivering drugs to the brain parenchyma. Understanding waste clearance in the brain is a vital field of study to be able to approach neurodeg enerative diseases and this study has produced a valuable tool in investigating perivascular structures. Within this project, Dr. Magdoom acquired the brightfield and laser confocal images. I designed the segmentation method and acquired the reconstructed images infusion and rat imaging together with the help of Julian Rey, a PhD candidate in the mechanical engineering department at UF. The tracer infusion was performed in th e lab of Michael A. King, who opened his histology lab to us. Figure 4 1. Correlation of micro vessel density (MVD) to K trans K trans was found to be positively correlated to MVD in VX2 tumors in rabbits. This image is F igure 2 from Li et al. [10]


42 Figure 4 2. Dynamic K trans study in VX2 tumor line rabbits. The onset of necrosis was found to occur one week post implantation. K trans was found to increase rapidly, then peak and fall off with the onset of necrosis. This image is F igure 1 from Li et al. [10] Figure 4 3. K trans and tumor clinical stage in human prostate cancer. K trans values were found to be higher in later clinical stages. This image is F igure 2a from Wu et al. [34]


43 Figure 4 4. Brightfield optical images of extracted brains. A) Nave rat, no tracer infusion. B) Rat with tracer (GD albumin localized to where the tracer is fi xed. C) The region of interest from B) magnified to show the lining of the tracer along the blood vessel. Dr. Magdoom generated these images.


44 Figure 4 5. the rat brain and liver following ex vivo MR imaging. A ), B ), and C) Tracer fluorescence imaged (red) merged with tissue autofluorescence image (green). D), E), and F) Tracer fluorescence images (red) merged with 1:1000 diamidino 2 phenylindole) image (blue). The presence of tracer causes fluorescence, and thus visible perivascular space in A ) B ) D ) E ) and F ) (shown with white arrows). The lack of tracer fluorescence in the liver C) confirms that the tracer is not bound to the luminal wall, and thus the tracer is localized to perivascular spaces in the brain. These images were obtained and generated by Dr. Magdoom.


45 Figure 4 6. Example of reconstructed perivascular network. Thresholding the T 1 weighted image to acquire the PVS network and overlaying the res ult onto the surrounded by perivascular space (blue) resulting from significant contrast enhancement relative to the rest of the original image. Figure 4 7. Further recons truction and extrapolation. Enhancement of the region of interest from Figure 4 6 reveals the perivascular network (blue) surrounding multiple cortical blood vessels in the brain (red).


46 LIST OF REFERENCES 1. Berger A (2002) Magnetic resonance imaging. BMJ 324: 35 2. Ballabh P, Braun A, Nedergaard M (2004) The blood brain barrier: an overview: Structure, regulation, and clinical implications. Neurobiology of Disease 16: 1 13 doi: 3. Jessen KR, Mirsky R (1980) Glial cells in the enteric nervous system contain glial fibrillary acidic protein. Nature 286: 736 737 doi:10.1038/286736a0 4. Fangusaro J (2012) Pediatric h igh grade glioma: a review and update on tumor clinical characteristics and biology. Front Oncol 2: 105 5. Johnson KE, Wilgus TA (2014) Vascular Endothelial Growth Factor and Angiogenesis in the Regulation of Cutaneous Wound Repair. Adv Wound Care (New Ro chelle) 3: 647 661 6. Nishida N, Yano H, Nishida T, Kamura T, Kojiro M (2006) Angiogenesis in cancer. Vasc Health Risk Manag 2: 213 219 7. Tong RT, Boucher Y, Kozin SV, Winkler F, Hicklin DJ, Jain RK (2004) Vascular normalization by vascular endothelial growth factor receptor 2 blockade induces a pressure gradient across the vasculature and improves drug penetration in tumors. Cancer Res. United States, pp 3731 3736 8. Tofts PS, Kermode AG (1991) Measurement of the blood brain barrier permeability and le akage space using dynamic MR imaging. 1. Fundamental concepts. Magn Reson Med 17: 357 367 9. Yeo DM, Oh SN, Jung CK, Lee MA, Oh ST, Rha SE, Jung SE, Byun JY, Gall P, Son Y (2015) Correlation of dynamic contrast enhanced MRI perfusion parameters with angio genesis and biologic aggressiveness of rectal cancer: Preliminary results. J Magn Reson Imaging 41: 474 480 doi:10.1002/jmri.24541 10. Li H, Yu L, Wang W, Wang L, Zheng X, Dai S, Sun Y (2018) Dynamics of angiogenesis and cellularity in rabbit VX2 tumors us ing contrast enhanced magnetic resonance imaging and diffusion weighted imaging. Oncol Lett 15: 2978 2984 11. Zhang N, Zhang L, Qiu B, Meng L, Wang X, Hou BL (2012) Correlation of volume transfer coefficient Ktrans with histopathologic grades of gliomas. J Magn Reson Imaging 36: 355 363 12. Zhao M, Guo LL, Huang N, Wu Q, Zhou L, Zhao H, Zhang J, Fu K (2017) Quantitative analysis of permeability for glioma grading using dynamic contrast enhanced magnetic resonance imaging. Oncol Lett 14: 5418 5426


47 13. Sub ashi E, Cordero FJ, Halvorson KG, Qi Y, Nouls JC, Becher OJ, Johnson GA (2016) Tumor location, but not H3.3K27M, significantly influences the blood brain barrier permeability in a genetic mouse model of pediatric high grade glioma. J Neurooncol 126: 243 25 1 14. Choi YS, Kim DW, Lee SK, Chang JH, Kang SG, Kim EH, Kim SH, Rim TH, Ahn SS (2015) The Added Prognostic Value of Preoperative Dynamic Contrast Enhanced MRI Histogram Analysis in Patients with Glioblastoma: Analysis of Overall and Progression Free Sur vival. AJNR Am J Neuroradiol. 2015 by American Journal of Neuroradiology., United States, pp 2235 2241 15. Wood JR, Green SB, Shapiro WR (1988) The prognostic importance of tumor size in malignant gliomas: a computed tomographic scan study by the Brain Tu mor Cooperative Group. J Clin Oncol 6: 338 343 doi:10.1200/jco.1988.6.2.338 16. Dong X (2018) Current Strategies for Brain Drug Delivery. Theranostics 8: 1481 1493 17. Jallo GI, Volkov A, Wong C, Carson BS, Sr., Penno MB (2006) A novel brainstem tumor mod el: functional and histopathological characterization. Childs Nerv Syst 22: 1519 1525 doi:10.1007/s00381 006 0174 8 18. Barth RF, Kaur B (2009) Rat brain tumor models in experimental neuro oncology: the C6, 9L, T9, RG2, F98, BT4C, RT 2 and CNS 1 gliomas. J Neurooncol 94: 299 312 19. Jones DK, Horsfield MA, Simmons A (1999) Optimal strategies for measuring diffusion in anisotropic systems by magnetic resonance imaging. Magn Reson Med. 1999 Wiley Liss, Inc., United States, pp 515 525 20. Weinmann HJ, Brasch RC, Press WR, Wesbey GE (1984) Characteristics of gadolinium DTPA complex: a potential NMR contrast agent. AJR Am J Roentgenol 142: 619 624 doi:10.2214/ajr.142.3.619 21. Chen X, Astary GW, Sepulveda H, Mareci TH, Sarntinoranont M (2008) Quantitative asses sment of macromolecular concentration during direct infusion into an agarose hydrogel phantom using contrast enhanced MRI. Magn Reson Imaging 26: 1433 1441 doi:10.1016/j.mri.2008.04.011 22. Rohrer M, Bauer H, Mintorovitch J, Requardt M, Weinmann HJ (2005) Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Invest Radiol 40: 715 724 23. Pickup S, Wood AK, Kundel HL (2005) Gadodiamide T1 relaxivity in brain tissue in vivo is lower than in saline. Magn Reso n Med 53: 35 40 doi:10.1002/mrm.20336 24. Ewing JR, Bagher Ebadian H (2013) Model selection in measures of vascular parameters using dynamic contrast enhanced MRI: experimental and clinical applications. NMR Biomed 26: 1028 1041 doi:10.1002/nbm.2996


48 25. Ha acke EM, Brown RW, Thompson MR, Venkatesan R (1999) Magnetic resonance imaging: physical principles and sequence design. Wiley Liss New York:, 26. Foottit C, Cron G, Mercier J, Nguyen V, Cameron I, Schweitzer M, Sinclair J, Woulfe J, Hogan M, Nguyen T Opt imizing perfusion imaging of brain tumors: Validation of venous output function used as a surrogate AIF. Proceedings of International Society for Magnetic Resonance in Medicine (ISMRM) 19th Annual Meeting, Montreal, Canada, p 2038 27. Bereczki D, Wei L, O tsuka T, Hans FJ, Acuff V, Patlak C, Fenstermacher J (1993) Hypercapnia slightly raises blood volume and sizably elevates flow velocity in brain microvessels. Am J Physiol 264: H1360 1369 28. Markwardt CB (2009) Non linear least squares fitting in IDL with MPFIT. arXiv preprint arXiv:09022850 29. Burnham KP, Anderson DR (2004) Multimodel inference: understanding AIC and BIC in model selection. Sociological methods & research 33: 261 304 30. Logan J, Fowler JS, Volkow ND, Wolf AP, Dewey SL, Schlyer DJ, MacGregor RR, Hitzemann R, Bendriem B, Gatley SJ, et al. (1990) Graphical analysis of reversible radioligand binding from time activity measurements applied to [N 11C methyl] ( ) cocaine PET stu dies in human subjects. J Cereb Blood Flow Metab 10: 740 747 doi:10.1038/jcbfm.1990.127 31. Wilcoxon F (1992) Individual comparisons by ranking methods. Breakthroughs in statistics. Springer, pp 196 202 32. Magdoom KN, Delgado F, Bohorquez AC, Brown AC, C arney PR, Rinaldi C, Mareci TH, Ewing JR, Sarntinoranont M (2019) Longitudinal evaluation of tumor microenvironment in rat focal brainstem glioma using diffusion and perfusion MRI. J Magn Reson Imaging 49: 1322 1332 doi:10.1002/jmri.26315 33. Louis DN, Ohg aki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A, Scheithauer BW, Kleihues P (2007) The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 114: 97 109 34. Wu X, Reinikainen P, Kapanen M, Vierikko T, Ryymin P, Kellokumpu Lehtinen PL (2018) Dynamic Contrast Enhanced Imaging as a Prognostic Tool in Early Diagnosis of Prostate Cancer: Correlation with PSA and Clinical Stage. Contrast Media Mol Imaging 2018: 3181258 35. Little RA, Barjat H, Hare JI, Jenner M, Watson Y, Cheung S, Holliday K, Zhang W, O'Connor JPB, Barry ST, Puri S, Parker GJM, Waterton JC (2018) Evaluation of dynamic contrast enhanced MRI biomarkers for stratified cancer medicine: How do permeability and perfusion vary between human tumours? Magn Reson Imaging 2017 The Authors. Published by Elsevier Inc, Netherlands, pp 98 105


49 36. Wu Q, Tyler B, Sukay L, Rhines L, DiMeco F, Clatterbuck RE, Guarnieri M, Carson BS, Sr. (2002) Experimental rodent models of brainstem tumors. Vet Pathol 39: 293 299 doi:10.1354/vp.3 9 3 293 37. Iliff JJ, Wang M, Liao Y, Plogg BA, Peng W, Gundersen GA, Benveniste H, Vates GE, Deane R, Goldman SA, Nagelhus EA, Nedergaard M (2012) A Paravascular Pathway Facilitates CSF Flow Through the Brain Parenchyma and the Clearance of 147ra111 doi:10.1126/scitranslmed.3003748 38. Xie L, Kang H, Xu Q, Chen MJ, Liao Y, Thiyagarajan M, O'Donnell J, Christensen DJ, Nicholson C, Iliff JJ, Takano T, Deane R, Nedergaard M (2013) Sleep drives metabolite clearance from the adult brain. Science 342: 373 377 39. Iliff JJ, Lee H, Yu M, Feng T, Logan J, Nedergaard M, Benveniste H (2013) Brain wide pathway for waste clearance captured by contrast enhanced MRI. J Cli n Invest 123: 1299 1309 40. O'Donnell J, Zeppenfeld D, McConnell E, Pena S, Nedergaard M (2012) Norepinephrine: a neuromodulator that boosts the function of multiple cell types to optimize CNS performance. Neurochem Res 37: 2496 2512 41. Wood TC, Lythgoe DJ, Williams SC rBET: making BET work for rodent brains. Proc Intl Soc Mag Reson Med, p 2706 42. Jenkinson M, Beckmann CF, Behrens TE, Woolrich MW, Smith SM (2012) FSL. Neuroimage. 2011 Elsevier Inc, United States, pp 782 790 43. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, Gerig G (2006) User guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. United States, pp 1116 1128 44. Swanson LW (2018) Brain maps 4.0 Struct ure of the rat brain: An open access atlas with global nervous system nomenclature ontology and flatmaps. J Comp Neurol 526: 935 943 45. Jenkinson M, Smith S (2001) A global optimisation method for robust affine registration of brain images. Med Image An al. Netherlands, pp 143 156 46. Jenkinson M, Bannister P, Brady M, Smith S (2002) Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage. United States, pp 825 841 47. Schindelin J, Arganda Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez JY, White DJ, Hartenstein V, Eliceiri K, Tomancak P, Cardona A (2012) Fiji: an open source platform for biological image analysis. Nat Methods 9: 676 682


50 48. Sato Y, Nakajima S, Shiraga N, Atsumi H, Yoshida S, Koller T, Gerig G, Kikinis R (1998) Three dimensional multi scale line filter for segmentation and visualization of curvilinear structures in medical images. Med Image Anal. Netherlands, pp 143 168


51 BIOGRAPHICAL SKETCH Alec majored in biochemistry and molecular biology. He completed his Associate of Arts degree at the University of Central Florida before transferring to the University of Florida to pursue new research opportunities. He completed a combined degree in which the undergraduate degree was a thesis based Bachelor of Science in interdisciplinary sciences with a concentration of biochemistry and molecular biology. His undergraduate thesis focused on imaging the perivascular network in the rat brain. Master of Science degree in summer of 2 019.