Title Page
 Table of Contents
 List of Tables
 List of Figures
 Review of literature
 Experimental equipment, hardware...
 Assessment of image quality: test...
 Evaluation of four conventional...
 Assessment of low-contrast detectability...
 Assessment of low-contrast detectability...
 Biographical sketch

Title: Objective assessment of image quality in conventional fluoroscopy and the low-contrast detectability of objects in carbon dioxide angiography
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00090912/00001
 Material Information
Title: Objective assessment of image quality in conventional fluoroscopy and the low-contrast detectability of objects in carbon dioxide angiography
Series Title: Objective assessment of image quality in conventional fluoroscopy and the low-contrast detectability of objects in carbon dioxide angiography
Physical Description: Book
Creator: Arreola, Manuel,
 Record Information
Bibliographic ID: UF00090912
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 001922338
oclc - 30495393

Table of Contents
    Title Page
        Page i
        Page ii
        Page iii
    Table of Contents
        Page iv
        Page v
        Page vi
        Page vii
        Page viii
    List of Tables
        Page ix
        Page x
    List of Figures
        Page xi
        Page xii
        Page xiii
        Page xiv
        Page xv
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
    Review of literature
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
    Experimental equipment, hardware and software
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
    Assessment of image quality: test procedures and experimental methodology
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
    Evaluation of four conventional fluoroscopy systems: results and discussion
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
        Page 153
        Page 154
        Page 155
        Page 156
        Page 157
        Page 158
        Page 159
        Page 160
    Assessment of low-contrast detectability in CO2 angiographic images: methodology
        Page 161
        Page 162
        Page 163
        Page 164
        Page 165
        Page 166
        Page 167
        Page 168
        Page 169
        Page 170
        Page 171
        Page 172
        Page 173
        Page 174
        Page 175
        Page 176
        Page 177
        Page 178
        Page 179
        Page 180
        Page 181
        Page 182
        Page 183
        Page 184
        Page 185
        Page 186
        Page 187
        Page 188
        Page 189
        Page 190
        Page 191
        Page 192
        Page 193
        Page 194
        Page 195
    Assessment of low-contrast detectability in CO2 angiographic images: results and discussion
        Page 196
        Page 197
        Page 198
        Page 199
        Page 200
        Page 201
        Page 202
        Page 203
        Page 204
        Page 205
        Page 206
        Page 207
        Page 208
        Page 209
        Page 210
        Page 211
        Page 212
        Page 213
        Page 214
        Page 215
        Page 216
        Page 217
        Page 218
        Page 219
        Page 220
        Page 221
        Page 222
        Page 223
        Page 224
        Page 225
        Page 226
        Page 227
        Page 228
        Page 229
        Page 230
        Page 231
        Page 232
        Page 233
        Page 234
        Page 235
        Page 236
        Page 237
        Page 238
        Page 239
        Page 240
        Page 241
        Page 242
        Page 243
        Page 244
        Page 245
        Page 246
        Page 247
        Page 248
        Page 249
        Page 250
        Page 251
        Page 252
        Page 253
        Page 254
        Page 255
        Page 256
        Page 257
        Page 258
        Page 259
        Page 260
        Page 261
        Page 262
        Page 263
        Page 264
        Page 265
        Page 266
        Page 267
        Page 268
        Page 269
        Page 270
        Page 271
        Page 272
        Page 273
        Page 274
        Page 275
        Page 276
        Page 277
        Page 278
        Page 279
        Page 280
        Page 281
        Page 282
        Page 283
        Page 284
        Page 285
        Page 286
        Page 287
        Page 288
        Page 289
    Biographical sketch
        Page 290
        Page 291
        Page 292
Full Text








It is with the deepest sense of gratitude that I want

to thank all the people that, in one way or another, have

contributed to the successful completion of this work. In

particular, two persons, with whom I have had the privilege

to work very closely: Dr. Lawrence T. Fitzgerald,

chairperson of my supervisory committee, and Dr. Libby

Brateman, cochairperson. My gratitude goes to Dr.

Fitzgerald for his guidance and support, and for constantly

sharing with me his vast experience in the field of medical

physics. My most sincere and heartfelt thankfulness goes to

Dr. Brateman, for her constant and outstanding guidance,

understanding, encouragement and moral support throughout

the last five years, and for giving me the highest sense of

ethics, humanity and respect for my profession; I will carry

that with me for the rest of my life. I am extremely

thankful to Dr. Irvin Hawkins, Dr. David Hintenlang and Dr.

Charles Roessler, members of my supervisory committee for

their guidance and understanding. I want to specially thank

Dr. William Properzio for attending the final defense of

this work on a very short notice.

I want to thank Mr. Edward Albert, from Medrad Inc.,

for providing the FluoroVision Contrast Enhancement System

used in this research at no cost during the extended period

of time needed for completion of the work. I want to thank

Mr. Peter Michael, from the Division of Bioengineering, for

his skillful manufacture of the low-contrast phantoms, and

Mr. Nikolaos Gkanatsios, fellow graduate student in charge

of the photographs included in this dissertation.

I want to extend my gratitude to Mrs. Beverly Hoyle and

Mr. Steve Knapp for five years of friendship and excellent

team work. They helped create a delightful working


Finally and most importantly, I want to thank my

parents, Ricardo and Elena Arreola, for providing me with a

solid childhood; my wife Ana, for her constant love,

encouragement, patience and understanding, without which I

could not have completed this work; and my sons, Daniel

Isaac and David Isaiah, for being my everlasting pride and


A last acknowledgement is for my native country,

Mexico, for shaping my ideology and my beliefs.




ACKNOWLEDGEMENTS . . . . . . . .. ii

LIST OF TABLES . . . . . . . . . . ix

LIST OF FIGURES . . . . . . . . ... . xi

ABSTRACT .. . . . . . . . . . xiv


1 INTRODUCTION . . . . . . . . 1

2 REVIEW OF LITERATURE . . . . . .. 10

2.1 Brightness Uniformity .. . . . 10
2.1.1 The NEMA Standard for Determination of
Integral Uniformity in Gamma Cameras . 11
2.1.2 Brightness Uniformity in Conventional
Fluoroscopy . . . . . . .. 13
2.2 Spatial Linearity . . . . . ... 15
2.3 System Contrast Ratio . . . . . 21
2.3.1 Traditional Definitions of Contrast
Ratio . .. . . . . . . 22
2.3.2 The NEMA Standard for Determination
of the System Contrast Ratio. .. ... 23
2.4 Conversion Efficiency ... . . . 27
2.5 Spatial Resolution . . . . . .. 28
2.5.1 Square-Wave Response Function Methods
and Limiting Spatial Resolution . . 29
2.5.2 Line Spread Function Method for
Determination of the Modulation Transfer
Function . . . . ....... .31
2.5.3 The Variance Method for MTF
Determination . . . . . . .. 32
2.6 Image Noise . ..... . . . . 33
2.7 Low-Contrast Detectability . . . .. 36
2.8 The Use of Carbon Dioxide in Digital
Subtraction Angiography . . . . .. 38



3.1 The FluoroVision(TM) Image Enhancement
System . . . . . . . . .. 44
3.1.1 Frame Grabbing and Digitization . 45
3.1.2 Image Enhancement and Noise Reduction 46
3.2 Communications Interface . .. . . 48
3.3 Dedicated Software . . . . . .. 50
3.4 Test Objects, Phantoms and Other Equipment 51
3.5 Fluoroscopic Systems Used . . . .. 53


4.1 Brightness Uniformity . . . . .. 59
4.1.1 Test Procedure for Calculation of
Integral Uniformity . . . . ... 59 Phantom . . . . . .. 60 Testing procedure . . . .. 60
4.1.2 Nonuniformity Corrections . . .. 64
4.2 Spatial Linearity . . . . . .. 65
4.2.1 Modifications to the Chakraborty Model
for an Objective Assessment of Spatial
Linearity . . . . . . ... .65
4.2.2 Test Procedure for Assessment of
Spatial Linearity . . . . .. 67 Phantom . . . . . . 67 Testing procedure . . . .. 70
4.3 System Contrast Ratio . . . . .. 78
4.3.1 Definition of System Contrast Ratio 79
4.3.2 Test Procedure for Determination of
the System Contrast Ratio . . . .. 81 Phantom . . . . . . 81 Testing procedure . . . .. 81
4.4 Relative Conversion Efficiency . . .. 83
4.4.1 Definition of Relative Conversion
Efficiency . . . . . . . 84
4.4.2 Test Procedure for Determination of
Relative Conversion Efficiency ..... 86 Phantom and equipment . . .. 86 Testing procedure ... . . 86
4.5 Modulation Transfer Function ...... 87
4.5.1 The Variance Method for Determination
of the MTF . . . . .. .. . 88
4.5.2 A Simplified Method for Determination
of the MTF Using the Variance Method . 91 Phantom . . . . . . 91 Testing procedure . . . .. 94


4.6 Image Noise . . . . . . . . 98
4.6.1 An Objective Assessment of Image
Noise . . . . . . . . 98
4.6.2 Objective Procedure for Assessment of
Image Noise . . . . . . .. 100 Phantom . . . . . . 100 Noise as a function of frame
averaging . . . . . . .. 100 Noise as a function of IER. . 102 Noise as a function of location
in image . . . . . . .. 103
4.7 Low-Contrast Detectability . . . .. 104
4.7.1 Low-Contrast Detectability in
Scatter-Free Conventional Fluoroscopy: A
Modification of the Rose Model . .. 107
4.7.2 A Quantitative Determination of
Contrast-Detail Curves in Conventional
Fluoroscopy . . . . . . . 110 Phantom . . . . . . 110 Testing procedure . . . .. 113


5.1 Brightness Uniformity . . . . .. 117
5.1.1 Determination of Integral Uniformity .117
5.1.2 Use of Correction Factors . . .. 119
5.2 Spatial Linearity . . . . . .. 125
5.2.1 Determination of Parameters . .. 126
5.2.2 Correction of Image Distortions. . 128
5.3 System Contrast Ratio . . . . .. 133
5.4 Relative Conversion Efficiency . . .. 135
5.5 Modulation Transfer Function . . .. 136
5.6 Image Noise . . . . . . .. . 140
5.6.1 Noise as a Function of Frame
Averaging . . . . . . . 141
5.6.2 Noise as a Function of IER . . .. 144
5.6.3 Noise as a Function of Location in
Image . . . . . . . . . 144
5.7 Low-Contrast Detectability . . . .. 156
5.8 Analysis of the Methodologies Implemented 159


6.1 Basic Description of the Angiographic
and DSA Imaging Processes . . . .. 162
6.2 Radiological Equivalence Between Air and
Carbon Dioxide . . . . . . .. 168
6.2.1 X-Ray Attenuation Characteristics . 168


6.2.2 Physical Properties . . . .. 169
6.3 Calculation of Subject Contrast in the
Presence of Scattered Radiation. . . .. .173
6.3.1 Subject Contrast with Scattered
Radiation . . . . . . . . 173
6.3.2 Determination of Scatter Degradation
Factors . . . . . . . . 174
6.4 Low-Contrast Detectability of Air-Filled
Objects in Angiographic Imaging . . . 180
6.5 Description of the Inserts for the DSA
Phantom . . . . . . . . . 182
6.5.1 The Air-Low-Contrast Insert (ALCI)
Phantom . . . . . . . . 183
6.5.2 The Air-Vessel Insert (AVI) Phantom .189
6.6 Experimental Procedure: Typical X-ray
Spectra and Anatomical Thicknesses . . 192

DISCUSSION . . . . . . . . 196

7.1 Characteristics of the Fluoroscopic System
Used . . . . . . . .... .. 197
7.1.1 X-Ray Spectra . . . . ... .197
7.1.2 Manual and Automatic Selection of
Fluoroscopic Factors . . . .. 203
7.2 Setup for Data Collection . . . .. 206
7.3 Calculated Scatter Degradation Factors
and Subject Contrasts . . . . .. 209
7.3.1 Scatter Degradation Factors ... .210
7.3.2 Subject Contrast . . . . .. 212
7.4 Study of the ABC Case with the ALCI Phantom 213
7.4.1 The Effect of Structural
Nonuniformities on the Low-Contrast
Detectability of Gas-Filled Objects . 214
7.4.2 Comparison of Contrast-Detail Curves 221
7.5 Study of the Manual kVp Case with the ALCI
Phantom . . . . . . . ... 223
7.5.1 The Lower Extremity Case . . .. .223
7.5.2 The Abdominal Case . . ... 227
7.6 Studies with the AVI Phantom .. . . 230

8 SUMMARY . . . . . . . . . .. 235
8.1 Objective Methods for Assessment of Image
Quality in Conventional Fluoroscopy . . 235
8.2 Low-Contrast Detectability of CO2
Angiographic Images . . . . . .. 240






UNITS EVALUATED . . . . . . . 268


THRESHOLD CONTRAST. . . . . .. .275

REFERENCES . . . . . . . . . . . 283

BIOGRAPHICAL SKETCH . . . . . . . . .. 290



Table page

2.1. Diameters of contrast ratio objects used in
system contrast ratio determination . . .. 26

3.1. Conventional fluoroscopy systems evaluated in
Shands Hospital at the University of Florida 54

4.1 Diameters and percent contrast values for
low-contrast objects in the LTO 10 . . ... 113

5.1. Brightness uniformity test: Calculated integral
uniformity (IU) for the four fluoroscopic
systems evaluated .. . . . . . . 118

5.2. Brightness uniformity test: Comparison of IU
before and after averaging and nonuniformity
correction factors CFij are used . . . .. 124

5.3. Spatial linearity test: number of wire
crossings from Leeds Test Object M1 used in
analysis of data: full-field mode . . .. .126

5.4. Spatial linearity test: calculated
characteristic parameters C, and C for the four
fluoroscopic systems evaluated: full-field mode 127

5.5. Spatial linearity test: calculated
characteristic parameters D. and D for the four
fluoroscopic systems evaluated: full-field mode 128

5.6. System contrast ratio test: calculated SCR for
the four fluoroscopic systems evaluated . . 133

5.7. Relative conversion efficiency test: calculated
RCE for the fluoroscopic systems evaluated . 135

5.8. Modulation transfer function: estimated
limiting spatial resolution calculated at
MTF = 0.1 for the fluoroscopic systems evaluated 140

Table page

6.1. Diameter (d) of the circular objects in the ALCI
phantom . . . . . . . . . . 188

6.2. Depths of circular objects in the ALCI phantom 188

6.3. Widths of grooves in the AVI phantom . . .. .189

6.4. Lucite Phantom thicknesses and kVp values
selected for the study of the low-contrast
detectability of carbon dioxide angiographic
images . . . . . . . . . . 193

7.1. Scatter degradation factors for the simulated
clinical cases of Table 6.4, interpolated from
data published by Honda et al. (Hon91) . . 211

7.2 Tube voltages (kVp) and currents (mA) utilized
in the ABC cases . . . . . . . 212

7.3 Percentage subject contrast for rows of objects
in the ALCI phantom: H = 12 cm and W = 23 cm 213

7.4 Percentage subject contrast for groups of
grooves in the AVI phantom: H = 12 cm and
W = 23 cm . . . . . . . . . 214


Figure page

2.1. The Chakraborty model for geometric distortion 17

4.1. Test equipment setup . . . . . .. 58

4.2. Effect of image averaging on random fluctuations 62

4.3. Fluoroscopic image of the Leeds Test Object Ml
obtained in fluoroscopic system A . . .. 69

4.4. Spatial linearity evaluation: effect of image
averaging . . . . . . . .. .. 73

4.5. Spatial linearity evaluation: effect of
applying corrections for image nonuniformities 75

4.6. Huttner Type 18 bar-pattern phantom .. ... 93

4.7. Example of proper positioning of the bar-pattern
phantom . . . . . . . .. .. . 96

4.8. Subdivision of the usable fluoroscopic field
into 529 square ROIs of dimensions 20x20 pixels 106

4.9. Layout of the Leeds Test Object 10 (LTO 10) 112

5.1. Brightness uniformity assessment: uncorrected
image . . . . . .. . . . . 121

5.2. Effect of the use of correction factors CFi, in
images of the uniform phantom obtained in system
D . . . . . . . . . .. . 123

5.3. Example of the use of the Chakraborty parameters
for spatial linearity corrections of the test
image of LTO Ml . . . . . . ... 130

5.4 Photographs corresponding to the example of the
use of the Chakraborty parameters for spatial
linearity corrections of Figure 5.3 . . .. .132


Figure page

5.5. MTF curves for three of the four fluoroscopic
systems evaluated . . . . . . .. 139

5.6. Image noise as a function of frame averaging 143

5.7. Image noise as a function of IER . .. ... 146

5.8. Image noise test results for system A . .. 148

5.9. Image noise test results for system B . . 150

5.10. Image noise test results for system C ... .152

5.11. Image noise test results for system D . . 154

5.12. Contrast-detail curves for the fluoroscopic
systems evaluated . . . . . .. ... 158

6.1. The basic angiographic procedure . . ... .164

6.2. Comparison of mass attenuation coefficients for
iodine, air and carbon dioxide as a function of
x-ray energy . ... . . . . ... 171

6.3. Imaging variables that determine the amount and
energy of scattered radiation . . . . 177

6.4. Basic components of the Nuclear Associates DSA
phantom . . . . . . . .. .. . 185

6.5. The air-low-contrast insert (ALCI) phantom . 187

6.6. The air-vessel insert (AVI) phantom .. ... 191

7.1. Characterization of the x-ray spectra used in
the study of low-contrast images in CO2
angiography. . . . . . . . . .. 200

7.2. Estimated x-ray spectra used in the study of CO2
angiographic images . . . . . . .. 202

7.3. Other characteristics of system B . .. ... 205

7.4. Experimental setup for collection of data in the
case of low-contrast detectability of CO2
angiographic images . . . . . . .. 208

7.5. Effect of structural nonuniformities on low-
contrast detectability in the ABC case ... 217


Figure page

7.6 Contrast-detail curves obtained after test
images of the ALCI phantom have been corrected
for structural nonuniformities . . . .. .220

7.7 Contrast-detail curves for the lower extremity
case . . . . . . . . .. . . 226

7.8. Contrast-detail curves for the abdominal case 229

7.9. Contrast-detail curves obtained with the AVI
phantom for both the abdominal and the lower
extremity cases . . . . . .. ... 232


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Manuel Arreola

August, 1993

Chairperson: Lawrence T. Fitzgerald, Ph.D.
Cochairperson: Libby Brateman, Ph.D.
Major Department: Nuclear Engineering Sciences

The use of carbon dioxide (CO2) as a contrast agent in

angiographic procedures has increased in recent years, as an

alternative for patients with coexistent renal and cardiac

dysfunction, because of its nontoxicity and low risk of

fatal reactions. However, the low x-ray attenuation of CO2

results in low-contrast signals of the C02-filled vessels.

In addition, inherent local variations in brightness and

noise in fluoroscopic images further reduce these contrast

signals, and the resultant image is often unusable for

diagnosis. As is the case for other image quality

parameters in fluoroscopy, traditional methods for the

assessment of low-contrast detectability make use of a

series of perceptive tests that yield subjective results,

which make the tasks of long-term monitoring and comparisons

among systems difficult to perform. A set of objective


methodologies that use quantitative, rather than perceived

information, is presented in this work. The methods make

use of digitized test images obtained using a frame grabber,

which are then analyzed using dedicated computer programs.

Brightness uniformity, spatial linearity, system contrast

ratio, conversion efficiency, modulation transfer function,

image noise and low-contrast detectability, were evaluated

in four fluoroscopic systems. The methodology for the

objective evaluation of low-contrast detectability was also

used for the case of CO2 angiography. For this purpose,

low-contrast phantoms with circular and vessel-like objects

were manufactured for use with an acrylic phantom for

patient simulation. Percentage threshold contrast values

(PTCs) were calculated using the Rose model modified for a

digital system in the presence of scattered radiation and

the simulated cases of lower extremity and abdominal exams.

Results show that PTCs in the periphery of the image are as

much as five times those in the central area. Frame

averaging and the correction of images for brightness

nonuniformities are shown to improve low-contrast

detectability in the periphery of the field to levels

comparable to those in the central region. Lower extremity

results indicate insignificant effects on PTCs from tube

voltages below 60 and the use of antiscatter grids.

Abdominal results show the need for low tube voltages and

antiscatter grids.



Since the time of its development in the late 1970s,

digital subtraction angiography (DSA) has become a widely

used radiological procedure. Both intravenous and

intraarterial DSA studies have made extensive use of

iodinated contrast agents. However, in recent years

intravenous DSA has been largely abandoned in most

facilities in the United States because of its poor image

quality as compared to intraarterial DSA (Rie82, Ver91).

The image quality of typical intraarterial DSA images has

been studied extensively over the years (Kru81, Coh82,

Rie82, Kru84). Significant advances and improvements in DSA

protocols and equipment, such as contrast injectors,

catheters, as well as dedicated hardware and software, have

caused this modality to become an important part of modern

radiology departments and a common interventional

radiological procedure.

For decades, it has been well-known that a significant

number of angiographic patients present histories of allergy

to iodinated compounds (She80, Wea91). Recently, in

searching for solutions to the toxicity and allergy problems

presented by iodine, I. F. Hawkins at the Department of

Radiology, University of Florida College of Medicine, has

successfully introduced carbon dioxide (CO2) as an

intraarterial contrast agent (Haw82, Haw85). To date, its

clinical use has shown excellent results, because of the

absence of patient complications observed and the low cost

of using CO2 as a contrast agent as compared to commonly

used present-day contrast agents (Haw82, Haw85, Wea91).

However, CO2 as a contrast agent provides lower subject

contrast than iodine-based contrast agents (Haw85, Wea91),

and an in-depth analysis to optimize the radiological

properties of the low-contrast images resulting from its use

in DSA has not yet been performed.

Low-contrast detectability can be defined as the

ability of a radiological imaging system to reproduce small

subject contrast differences, such as those found in

intraarterial-carbon-dioxide DSA. The detectability of low-

contrast objects is known to be strongly limited by the

total noise content of the image (Har86). Thus, a reliable

evaluation of image noise and low-contrast detectability is

fundamental in the study of intraarterial carbon dioxide DSA

images. The evaluation of low-contrast detectability is an

important part in the assessment of image quality not only

for DSA systems, but also for digital and conventional

fluoroscopy (Lin82, Hig85, Oha86, Cow87). In addition to

low-contrast detectability and image noise, typical methods

for the assessment of image quality in conventional

fluoroscopy require performing tests that evaluate other

important characteristics of fluoroscopic systems, such as

brightness uniformity, spatial linearity and spatial

resolution (Lin82, Cow87).

Because of the analog nature of conventional

fluoroscopy, image quality parameters have been

traditionally evaluated making use of qualitative approaches

which restrict the assessment of image quality to a series

of perceptive determinations that yield subjective '

information. In general, this information cannot be used to

produce a reliable assessment of the performance of

conventional fluoroscopy systems unless a drastic change in

the quality of the image has occurred, and long-term

monitoring and intersystem comparisons depend heavily on the

subjective responses of individual observers.

Since the observer's visual perception is used as the

instrument to extract information from test images in

traditional methods, the results obtained from subjective

information are heavily influenced by several factors.

Among others, the visual limitations of the observer, as

well as the observation and ambient light conditions during

testing significantly affect the outcome of the tests.

However, the most significant effect comes from

discrepancies among different observers, because different

observers typically arrive at different results. This

inconsistency is a result of the differences in the visual

acuity among different observers as well as of the different

decision processes used by various observers in the analysis

of a test image (Mye85, Oha86).

Other image quality and performance parameters, such as

contrast ratio and conversion efficiency, can be determined

in a more quantitative manner (Moo81, Bro82, Nat86, Nat92).

However, standard procedures for their measurement are not

designed to be implemented in the clinical setting, and they

are difficult, if not impossible, to implement in the

clinical setting.

Thus, the assessment of image quality in conventional

fluoroscopy presents two serious problems: first, that of

the subjectivity in the determination of image quality

parameters, such as spatial resolution and low-contrast

detectability, and second, that of the difficult

implementation of certain testing procedures, such as those

for conversion efficiency and contrast ratio. These

problems show that a need exists for the development of a

simple methodology for the objective assessment of image

quality in fluoroscopic systems in general. In particular,

the study of the low-contrast detectability of CO2-filled

objects in DSA requires the development of a quantitative,

objective method for low-contrast detectability assessment

that is not influenced by limitations of an observer or

discrepancies among observers.


In recent years, typical nuclear medicine gamma camera

systems have incorporated certain objective image quality

protocols along with corrective software for brightness

nonuniformities and distortions (Mue81). Although DSA and

digital fluoroscopy systems have not yet incorporated

similar features, their digital capabilities have made it

possible to obtain objective information from fluoroscopic


Because of this, some of the traditional methodology

for image quality assessment in conventional fluoroscopy can

be adapted in a simplistic way for its use in DSA and

digital fluoroscopy systems. More important, though, is the

fact that such modified methodology can be used in

conventional fluoroscopy systems as well, with the simple

addition of a device to capture and digitize the analog

fluoroscopic image. With such an addition, a standard

methodology for objective image quality evaluation and image

correction can be implemented for general use in

conventional and digital fluoroscopy, as well as DSA


The basic research objectives of this work, which

served as primordial guidelines in the development of such

quantitative methodologies, can be listed as follows:

A. To develop a set of simple, reproducible

methodologies for the objective assessment of image quality

in conventional and digital fluoroscopy. A detailed

description of the procedure for data collection, the test

objects and phantoms used, and the appropriate algorithms

and computer programs for data analysis is indispensable for

each of these methodologies. In particular, a test for

assessment of low-contrast detectability is required as part

of the set.

B. To implement the methodologies developed by

performing initial objective evaluations of conventional

fluoroscopy units, and propose an initial interpretation of

the objective results obtained.

C. To modify the methodology for the assessment of

low-contrast detectability indicated in the first objective,

for the simulated clinical case of CO2 angiography,

including the effects of scattered radiation and of the low

relative attenuation of x-rays by CO2.

D. To use such modified methodology for the evaluation

of the characteristics of low-contrast CO2 images in two

simulated clinical cases, namely those of lower extremity

and abdominal examinations.

In accomplishing these objectives, a basic premise of

creating a set of simple, reproducible test procedures that

provide objective, quantitative information about the

quality of the image in a fluoroscopic system, as well as

the evaluation of low-contrast detectability characteristics

of CO2 angiographic images, was followed. More

specifically, these basic objectives of the work have been

achieved in the following manner:

a. Methodologies were developed to perform a thorough

evaluation of the image chain system by including procedures

for evaluation of the principal image quality parameters,

including (Lin82, Cow87):

i. Brightness uniformity.

ii. Spatial linearity.

iii. System contrast ratio.

iv. Conversion efficiency.

v. Image noise.

vi. Spatial Resolution.

vii. Low-contrast detectability.

b. Simple approaches to the assessment of these

various image quality parameters that produce results that

are easy to interpret were either adapted from previously

published methods or completely developed in an effort to

maintain the basic premise of simplicity described above.

c. In designing simple test procedures for image

quality assessment, these were designed to make use of

phantoms and test objects commercially available and

typically used in the clinical setting for the evaluation of

fluoroscopic systems.

d. In particular, the procedure for assessment of low-

contrast detectability was developed for easy implementation

in the case of scattered radiation present, in simulating

the case of CO2-filled objects in angiography.

The methodology developed for the purposes described

above is described in this dissertation. The theoretical

structure of this methodology, along with its practical

implementation and applications, not only to conventional

fluoroscopy systems, but in particular to the experimental

evaluation of low-contrast detectability in intraarterial

CO2 angiography, are presented. Finally, a detailed

analysis of the results obtained is included, and the

perspectives for future research and practical

implementation of these methodologies are discussed.

This document comprises two major portions: the first

five chapters are focused on the assessment of image quality

in conventional fluoroscopy. Chapter 2 contains a

literature review, including definitions of the image

quality parameters included in the methodology, along with

descriptions of traditional methods for their assessment, as

well as a review of literature on the use of CO2 in

angiography. Chapter 3 describes hardware and software as

the necessary tools for the procedures developed. Detailed

descriptions of the test procedures, the phantoms and test

objects utilized are included in Chapter 4. Results

obtained from the implementation of the objective

methodologies in four conventional fluoroscopic systems in

routine use in the Department of Radiology in Shands

Hospital at the University of Florida are presented in

Chapter 5.

Starting with the procedure for low-contrast

detectability described in Chapter 4, Chapter 6 describes

elementary angiographic techniques and the use of CO2 in

angiographic procedures. It also includes the description

of the experimental methodology followed in the study of the

low-contrast detectability of C02-filled objects in DSA, for

the cases of lower-extremity studies as well as abdominal

studies. The phantoms specifically designed for these

purposes are described in the latter sections of that

chapter. Results obtained from the experimental procedure,

are presented and discussed in Chapter 7. Chapter 8

includes general conclusions and considers future work

related to the document. Detailed descriptions of the

dedicated software are presented in Appendices A and B, and

technical information about hardware and equipment used in

this research are included in Appendix C.


Methods for the assessment of image quality in

conventional fluoroscopy performed in the clinical setting

have evolved over the years in order to provide more

reliable and useful information. A brief historical review

of the evolution of these methods and the results they

provide is presented in this chapter. The image quality

parameters described include: brightness uniformity,

spatial linearity, contrast ratio, conversion efficiency,

limiting spatial resolution, modulation transfer function,

image noise and low-contrast detectability.

2.1 Brightness Uniformity

The presence of brightness nonuniformities (i.e.,

inherent variations in brightness between adjacent regions

of medical diagnostic images) was first studied by various

authors for the cases of nuclear medicine (Mue81, Hug88) and

magnetic resonance imaging (MRI) (Bra86, Con87, Pri90). In

fact, the concept of brightness uniformity as an image

quality performance parameter was originally developed and

standardized for nuclear medicine gamma cameras by the

National Electrical Manufacturers Association (NEMA) by

defining two characteristic uniformity parameters (Mue81,

Nat86): integral uniformity and differential uniformity.

Of these two, integral uniformity is of particular

significance since it provides means to characterize

brightness uniformity quantitatively with a single

descriptor by following the simple methodology described in

the section below.

2.1.1 The NEMA Standard for Determination of Integral
Uniformity in Gamma Cameras

The currently used NEMA standard for nuclear medicine

gamma cameras establishes the following protocol for

determination of integral uniformity (Mue81, Nat86):

1. A flood-field image is obtained using a source of

99'Tc located at a distance of at least five times the

diameter of the useful field of view of the camera. A

minimum of 4000 counts is collected in each of the 64x64

digital image matrix elements.

2. In order to reduce random variations, the data are

smoothed once using a nine-point mask with the following


1 2 1

2 4 2

1 2 1


3. Integral uniformity is determined by searching for

the elements with the maximum and minimum number of counts

in the image matrix and by applying the definition

INTEGRAL UNIFORMITY = 100 ( Max Min ) (2.1)
( Max + Min )

where Max and Min correspond to the maximum and minimum

number of counts found among the elements of the smoothed

image matrix, respectively.

Different brightness uniformity parameters have been

proposed and used by others. Hughes and Sharp (Hug88) have

made use of the coefficient of variation of the counts per

image matrix element instead of integral uniformity. In

place of differential uniformity, they have used the spread

of differential uniformity, which is a measure of the width

of the frequency distribution of the differential uniformity

between pixels in a row or a column. Both of these

quantities involve numerical calculations far more

complicated than those involved in the determination of

integral and differential uniformity. To date, however, no

other standard for brightness uniformity has yet been

adopted in clinical nuclear medicine.

Brightness uniformity has also been identified as an

important image quality parameter in MRI, and its extent and

effect on clinical images has been previously studied

(Con87). Brateman et al. (Bra86) evaluated uniformity using

a simple phantom consisting of polyethylene reservoirs

filled with manganese chloride in saline solution. More

recently, the American Association of Physicists in Medicine

(AAPM) has adapted some of the standardized definitions for

integral uniformity (as used for gamma cameras) for MRI

medical systems (Pri90).

2.1.2 Brightness Uniformity in Conventional Fluoroscopy

In the case of conventional fluoroscopy, brightness

nonuniformities arise mainly from four sources:

1. Spatial nonuniformities inherent to the focusing

electromagnetic fields of the image intensifier (Cso85);

2. Aberrations in the optical coupling between the

image intensifier and the television (TV) camera (Boo91);

3. Vignetting in the video signal (Ver91); and

4. Inherently different x-ray absorption

characteristics for different regions of the image

intensifier input phosphor (Swa73). Such differences result

from local, slight variations in the composition and

thickness of the phosphor coating at the input surface.

These factors affect the local absorption of incident x rays

and propagate light unequally to the photocathode in various

regions of the input phosphor (Swa73), which results in

nonuniform brightness levels both at the input phosphor and

in the fluoroscopic image.

Despite having these sources identified, quantitative

evaluation and/or correction algorithms for brightness

nonuniformities have never become a part of image quality

assessment protocols for conventional fluoroscopy systems.

Recently, Cooney et al. (Coo88) suggested that a modified

definition of integral and differential uniformity, as

defined for gamma cameras and MRI systems, could be used as

an objective image quality parameter for digital

fluoroscopy. However, integral uniformity has not been

routinely employed as a quantitative image quality parameter

and therefore its determination has not been standardized

for the purposes of conventional fluoroscopy systems.

At present, algorithms for correction of brightness

nonuniformities are routinely used as part of the daily

quality control procedure followed for gamma cameras.

Different algorithms have been published and used for this

purpose. Sorenson and Phelps (Sor87) have reviewed some of

the most commonly used algorithms. In the case of MRI

systems, appropriate correction algorithms have also been

developed and used (Con87). Boone et al. (Boo91) have

suggested the use and implementation of correction

algorithms for nonuniformities in digital fluoroscopy.

However, no professional or scientific association has yet

adopted an algorithm for the correction of brightness

nonuniformities in digital or conventional fluoroscopy as a


2.2 Spatial Linearity

Spatial linearity is an image quality parameter defined

to evaluate the amount of geometric distortion in a medical

diagnostic image. The term refers to the altered geometric

appearance of the image with respect to that of the object

being imaged. The assessment of spatial linearity in a

conventional fluoroscopy system has been studied since the

early 1960s, when the first electromagnetically-focused

image intensifiers were developed, and several different

approaches to the problem have been published since then.

In the particular case of fluoroscopy, the primary

source of image distortion arises from the nearly spherical

input phosphor surface required for proper electromagnetic

focusing (Cso85). This distortion, known as pincushion

distortion, is purely geometric in nature (Cha87) and

results in a dissimilar mapping of points (x,y) on the

object plane into points (x',y') at the image surface in the

image intensifier input phosphor, as shown in Figure 2.1.

Using simple geometry in the typical fluoroscopy

situation depicted in Figure 2.1, it is possible to see that

the segment AB on the object plane produces a projected

image segment of length CD on the image plane. Images in

that plane are thus magnified with respect to the objects by

a position-dependent magnification factor M(x,y) given by

Figure 2.1. The Chakraborty model for geometric distortion
(adapted from Cha87).









f0C C


M(x,y) = = SID (2.2)
AB ( SID -h )

where the SID (source-to-image receptor distance), h and d

are defined in Figure 2.1. M is dependent on (x,y), because

d takes different values which depend on the object

coordinates. Because the radius of curvature of the image

intensifier input surface is usually unknown and impossible

to measure in the clinical setting, it is uncommon to have

precise knowledge of d for different (x,y). Because of the

dependence of M on (x,y) and h, different segments in the

same object plane are magnified differently, which results

in pincushion distortion in the image (x',y').

In addition to geometric sources, nongeometric sources

also contribute to distortion in the fluoroscopic image.

The most significant of these is the distortion due to

external magnetic fields, known as s-shaped distortion,

caused mainly by the Earth's magnetic field and noticeably

accentuated by stray magnetic fields (Cow87). Other minor

sources of nongeometric distortion include nonuniformities

resulting from electron optics in the image intensifier,

distortions in the optical coupler, TV camera and, for

digital fluoroscopy systems, imperfections in the analog-to-

digital converter (ADC) device, as indicated by Chakraborty


Spatial linearity is evaluated by determining the

difference between the position of a point (x,y) in the

object and the corresponding location (x',y') in the image

matrix, both of them relative to the actual center of the

fluoroscopic image.

Different parameters have been proposed to quantify

spatial linearity in a conventional fluoroscopy system.

Casperson (Cas76) proposed a model that accounts only for

geometric distortion and defines a distortion parameter

which is a function of the radial distance from the center

of the image. Though simple in its approach, the model

fails to take into account nongeometric contributions to

image distortion, which makes it incomplete when it is

necessary to include all sources of distortion in the

assessment of spatial linearity.

The model developed by Chakraborty (Cha87) defines a

simplistic set of parameters to evaluate spatial linearity

which takes into account both geometric and nongeometric

contributions and also incorporates an algorithm for

correction of image nonlinearities. This method is

explained in detail in the following paragraphs.

The Chakraborty model for assessment of spatial

linearity. In 1987, Chakraborty (Cha87) presented a simple

approach for the assessment of spatial linearity in

conventional fluoroscopy units which analyzes the geometric

relations among the x-ray source, the object and its image.

As the image is first formed in the near-spherical

input phosphor of the image intensifier, geometric

(pincushion) distortion occurs. Using the notation

convention for object and image coordinates described in

Figure 2.1, the relationship between object and image

coordinates at this first stage of the imaging chain is

given by the following equations:

x = x M(x,y) (2.3)

S= y M(x,y) (2.4)

where M(x,y) is the position-dependent magnification factor

defined in equation 2.2.

The Chakraborty model accounts also for nongeometric

distortions with the assumption that no second-order

distortions occur at any of the other stages in the imaging

chain. Thus, the model assumes a linear relationship

between the image coordinates at the image intensifier input

phosphor and the coordinates (pixel locations) in the

digital image (x",y") as follows:

x// = Ax x/ + Dx (2.5)

y// = Ay y + Dy


where Ax, A Dx and D are empirical constants which are

specific for a given fluoroscopic unit and a particular

image intensifier magnified mode used, with their values

determined experimentally. Thus, according to this model,

it is possible to characterize spatial linearity from both

geometric and nongeometric sources by determination of the

parameters in equations 2.3, 2.4, 2.5, and 2.6.

Other authors have investigated image distortion from

other perspectives (Boo91). To date, however, none of these

methods has been adopted as a standard protocol for the

routine assessment of spatial linearity in conventional


2.3 System Contrast Ratio

An evaluation of the ability of a conventional

fluoroscopy system to image differences in subject contrast

and accurately display them as similar differences in image

contrast (and thus, their corresponding gray shades) is

obtained through determination of the contrast ratio of the

system. Veiling glare, i.e., the effect of scattered

radiation and the spread of light within the output of the

image intensifier and optical coupler, is the main cause for

contrast degradation in fluoroscopic images (Sei85).

2.3.1 Traditional Definitions of Contrast Ratio

The quantitative determination of contrast ratio is a

common practice in the manufacturing process of image

intensifier tubes; however, in the clinical setting it is

impossible to reproduce the testing conditions used in

laboratory measurements because it is not possible to

uncouple the image intensifier from other components of the

imaging chain (Hen85). To overcome this difficulty,

traditional methodologies have used a variety of approaches

for use in the clinical environment. Common procedures

(Bro82) involve performing light intensity measurements at

the image intensifier output for two cases: that of an

image in which the central region of the image intensifier

input is blocked by a lead disk of standard dimensions, and

that of an image in which the image intensifier input is not

blocked. Contrast ratio is simply calculated as the

following quotient:



BND= Illuminance measured from image without the lead disk

(lux or footcandles),

BD= Illuminance measured from image with the lead disk (lux

or footcandles) and

BBK= Background illuminance (lux or footcandles).

Thus, determination of the contrast ratio for image

intensifier tubes by this method requires the use of a light

intensity meter (Bro82) and a suitable probe.

Other approaches make use of radiographic methods. In

particular, Moore (Moo81) proposes the use of a small format

camera commonly found in some fluoroscopic units. In this

method, a strip of lead is centered in the fluoroscopic

field and a small-format radiograph obtained. Therefore,

data for determination of the contrast ratio are obtained

from optical density instead of illuminance measurements in

the dark and light regions of the radiograph.

All of these conventional methods for determination of

contrast ratio involve measurements that are difficult to

perform and reproduce, and the reliability of the results

obtained from their implementation is therefore limited.

Routine image quality assessment requires a simpler, more

objective method for the determination of contrast ratio.

2.3.2 The NEMA Standard for Determination of the System
Contrast Ratio

In recent years, the wide variety of approaches

reported in the specialized literature for the measurement

and determination of the contrast ratio has made comparisons

among fluoroscopic systems a rather difficult task. This is

particularly true not only as a result of the different

methodologies that can be used, but also because of the

fact that contrast ratio can be measured at various stages

of the imaging chain in fluoroscopic systems.

The recently published NEMA standard for the

determination of contrast ratio (Nat92) provides a way to

give validity to such comparisons by modifying the

traditional definition of contrast ratio, which was strictly

reserved for the image intensifier tube, to that of a

"system contrast ratio" for the whole imaging chain.

According to the standardized procedure, this parameter can

be determined using data obtained at any stage of the

imaging chain. This versatility, however, is overshadowed

by the fact that the standard is applicable only to systems

in which the fluoroscopic technique factors, kVp and mA, are

controlled by the operator and are independent of each other

(Nat92). Thus, in a strict sense, the standard has a very

reduced practical applicability, because it cannot be used

in a very significant proportion of all fluoroscopic

systems: those in which the technique factors (kVp and mA)

are automatically determined by the automatic brightness

control (ABC), and those in which, although it is possible

to select kVp values, the system still determines the

corresponding mA.

In the case of fluoroscopy, where the brightness of the

images at the surface of the video monitor screen is used to

determine the system contrast ratio, the standard procedure

includes the following steps:

1. With the x-ray source off, the brightness control

of the video monitor is adjusted such that the TV raster

lines are just visible.

2. A blank image is obtained using an x-ray beam of 75

kVp and no object in the fluoroscopic field. The mA value

used for this image is that which is automatically selected

by the fluoroscopic system. Using a photometer with a

sampling aperture no larger than 5 mm, the contrast control

of the video monitor is increased to produce a brightness of

30 ft lambert at the surface of the monitor screen.

3. A second image is obtained by placing a lead disk

of standard dimensions in the fluoroscopic field and

maintaining the same values of tube potential (75 kVp) and

current (mA) used in step 2. To achieve this, it must be

possible for the operator to disable the automatic

brightness control and select the desired kVp and mA values

manually. For the image thus obtained, the corresponding

monitor screen luminance L and input exposure rate 12, as

defined in Chapter 4, are measured.

The standard lead disk, also called a "contrast ratio

object" must be 3 mm thick and have a diameter equal to 10%

of the nominal diameter of the fluoroscopic field being

evaluated. Table 2.1 presents the corresponding standard

dimensions of contrast ratio objects for typical nominal

diameters of image intensifiers.



11.4 (4.5") 3.6
15.2 (6") 4.8
17.8 (7") 5.6
22.9 (9") 7.2
25.4 (10") 8.0
30.5 (12") 9.6
35.6 (14") 11.2

4. A third image is obtained by removing the contrast

ratio object from the fluoroscopic field. Keeping the x-ray

tube potential constant at 75 kVp, the tube current (mA) is

then reduced until the brightness at the center of the

monitor screen is equal to that of the second image (L).

The corresponding input exposure rate Ii is measured. The

system contrast ratio (SCR) is then calculated as



Because of its limited applicability, as explained in

paragraphs above, the procedure in the NEMA standard was not

considered as part of the set of objective tests for the


assessment of image quality in fluoroscopy. Some specifics

of the standard, however, such as the test images used and

the dimensions of contrast ratio objects, were adapted for

the methodology described in section 4.3.

2.4 Conversion Efficiency

The measurement of the efficiency with which an image

intensifier converts a standard x-ray exposure rate at the

input phosphor into a certain image brightness at the output

phosphor has been standardized by the International

Commission on Radiation Units and Measurements (ICRU) as the

conversion efficiency (Int62, Nat79). Conversion efficiency

is defined as the ratio of the luminance at the image

intensifier output (in cd/m2), to the exposure rate at the

input phosphor (in mR/sec). The ICRU standard specifies the

type of x-ray beam to be used in measuring this parameter.

The very detailed specifications of the procedure make the

implementation of this standard in the clinical setting

extremely difficult, if not impossible (Hen85).

A more accessible approach to this problem that can be

incorporated into routine image quality assessment is

defined as a modified conversion efficiency. The procedure

for its determination usually consists of measuring only the

input exposure needed to produce a certain level of the

output video signal (Mis79). Because data are obtained at

the output video stage, the modified conversion efficiency


accounts for all contributors along the imaging chain. The

modified conversion efficiency can thus be used to monitor

the long-term stability of a particular image intensifier or

for comparison between units in an easier and more reliable

manner than using conversion efficiency. It presents,

however, similar difficulties in performance of measurements

and nonreproducible results as traditional contrast ratio

measurements, as explained in Section 2.3 (Hen85), and a

different approach to this problem is needed for

implementation of modified conversion efficiency

measurements in routine image quality assessment.

2.5 Spatial Resolution

The ability of an imaging system to reproduce fine

object detail in an image is evaluated by determining the

spatial resolution of the system. Spatial resolution is

assessed by measuring the response of an imaging system to

various spatial frequencies of interest. Numerous

techniques have been employed for this assessment in

conventional and digital fluoroscopy, including

determination of: the square-wave response function; the

limiting spatial resolution; and the line spread function,

or its Fourier transform, the modulation transfer function

(MTF). The MTF represents the ability of an imaging system

to reproduce contrast as a function of spatial frequency,

and its measurement has been studied for all types of


radiological modalities (e.g., radiography, fluoroscopy, CT)

as well as for the various components of such systems

(Col54, Jud76, Dro85, Hen85, Cun87, Sor87, Pri90).

Typically, the MTF is obtained using analytical

methods. Nonetheless, it can also be measured directly,

with the use of a phantom of sine-wave patterns of various

spatial frequencies. However, the construction of such a

phantom is extremely difficult, and a practical and commonly

used approximation to it is found in a square-wave phantom.

Such objects are known as high-contrast-resolution or bar-

pattern phantoms; typical bar-pattern phantoms for image

quality evaluations consist of bar patterns with spatial

frequencies that range from 0.25 to 10.0 Ip/mm.

Different methods can be used for the assessment of

spatial resolution. A brief review of some of the most

commonly used methods is presented in the following


2.5.1 Square-Wave Response Function Methods and Limiting
Spatial Resolution

The square-wave response function describes the

response of an imaging system to a square-wave input. In

the fundamental square-wave response method, originally

developed for radiographic systems by Coltman (Col54), an

image of an appropriate bar-pattern phantom is recorded on

film and then sampled with the use of a scanning


microdensitometer. The optical density at each frequency is

converted into a modulated amplitude represented by the

relative exposure using the characteristic curve of the

imaging system in question, and the system frequency

response is determined from these relative exposures. The

MTF of the system can be obtained by transforming the

square-wave response into a sine-wave response function

(Col54). Cunningham and Fenster (Cun87) have incorporated

this method for its use in dedicated digital fluoroscopy

systems by obtaining a digital fluoroscopic image of a bar-

pattern phantom and using pixel values instead of optical

densities as the parameter to transform into relative

exposures for determination of the modulated amplitudes at

each spatial frequency.

With this method, determination of the imaging system

contrast response to various frequencies requires conversion

into relative exposure, which is time-consuming and

demanding. Frequently, assessment of the limiting spatial

resolution of an imaging system is sufficient. The limiting

spatial resolution is obtained from a subjective

determination of the highest spatial frequency visually

resolved by an observer. The limiting spatial resolution

has been related as the spatial frequency at which the MTF

has dropped to 10% (Jud76). In comparing different systems,

the limiting spatial resolution may not be the comparison

method of choice, because one system may have a higher

limiting resolution while its contrast response at lower

frequencies may be poorer and more significant during

clinical use of the system.

2.5.2 Line Spread Function Method for Determination of the
Modulation Transfer Function

The line spread function (LSF) is the response of an

imaging system to a one-dimensional impulse. It may be

measured directly in analog or digital form, or it may be

obtained from differentiation of the edge spread function

(ESF) of the system, i.e., the response to a one-dimensional

edge, as in the original method introduced by Judy (Jud76)

for determination of the MTF of CT systems.

In the case of digital fluoroscopy, a common practice

to obtain the LSF directly is to evaluate the system

response to a narrow slit (Fuj85). This practice might

present severe problems with aliasing and undersampling, but

these problems can be overcome using the method proposed by

Sones and Barnes (Son84) for digital fluoroscopy systems.

In this method, the shift in spatial frequency caused by

undersampling is overcome by using fine wires slightly

angled with respect to the image matrix. The method

proposed by Judy (Jud76) for obtaining the ESF of a CT

system has also been modified for implementation in digital

fluoroscopy systems. A modification of this procedure by

Cunningham and Fenster (Cun87) introduces a method in which

numerical calculations involved in Fourier analysis are

greatly reduced by obtaining the LSF through finite-element

differentiation of the system ESF.

Once the LSF has been either measured directly or

determined analytically, the MTF is simply obtained as the

Fourier transform of the LSF.

2.5.3 The Variance Method for MTF Determination

Another method for determination of the MTF, especially

remarkable for its simplicity, was originally introduced by

Droege and Morin for CT systems (Dro82). In this procedure,

the image of a bar-pattern phantom is obtained, and regions

of interest (ROIs) are selected within each spatial

frequency bar-pattern in the phantom. Using the pixel

values within each ROI, the modulated amplitude at a

particular frequency is obtained from calculating the

variance of the pixel values in that particular ROI. The

MTF is then obtained using a simplification of Coltman's

method (Col54)(1). Droege and Rzeszotarski (Dro85) adopted

this method for the case of a digital fluoroscopy system.

The method is immune to aliasing, a major difficulty

encountered in the implementation of LSF methods for digital

systems, and it does not have the limitation of the square-

wave response method in requiring knowledge of the

characteristic curve of the imaging system. Such features

(1) The variance method is described in detail in Chapter 4.

make this method a more convenient and simpler choice to

determine the MTF of a digital imaging system.

2.6 Image Noise

The term image noise refers to the spatially dependent

variation in image intensity, or mottle, observed in any

image. Because of its consequences in medical diagnosis, it

becomes of significant clinical importance in medical

diagnostic imaging. The sources of image noise and the

impact that they have in the overall image quality of any

imaging system have been extensively studied for all

diagnostic imaging modalities over the years. In the case

of fluoroscopy, noise arises from both deterministic (or

structural) and random sources (Kru84). Structural noise

arises from various causes, such as (Coh82):

Vignetting in the image intensifier and the optical

components of the imaging chain;

Imperfections and nonuniformities in the image

intensifier input and output surfaces, such as surface

curvatures and nonuniform phosphor coating;

Nonuniformities in the TV input surface, such as

photoconductor blemishes.

Nonstructural noise arises mainly from three sources


The inherent quantum nature of x-ray photons, thus

making it dependent on exposure rate.

The electronic noise inherent to the TV system.

Digitization noise, also known as analog-to-digital

converter (ADC) noise.

These sources, and the importance and influence that

image noise have on image quality in fluoroscopy, have been

studied by various authors for digital fluoroscopy systems

since the years following the development of digital

subtraction angiography (DSA) systems. In particular,

Kruger, Mistretta and Riederer, and Harrison (Kru81, Har86)

have identified and evaluated such sources in the case of

digital fluoroscopy. Such evaluation establishes that the

total image noise at in a digital fluoroscopy system can be

quantitatively expressed as follows (Kru81, Har86):

a 2 = (B b G )2 N + B e (2.9)


N = Number of x-ray photons per pixel area detected at the

image intensifier input surface,

G = Conversion factor from x-ray flux at the image

intensifier input to light photon flux at the TV

photoconductor; this factor includes image intensifier

gain, iris aperture setting and loss of light intensity

due to the optical coupler,

b = Conversion factor from light at the TV photoconductor

to video output signal,


Oe = Constant additive electronic noise associated with the

video signal; and

B = Conversion factor from video signal to pixel value.

In the early years of fluoroscopy, the determination of

the amount of noise present in a fluoroscopic image began

with subjective evaluations of simple observations. One

current method in common use subjectively classifies a

fluoroscopic image into one of several descriptive

categories, which range from very noisy, to noisy, to noise-

free images. Other methods, such as determining noise

levels from voltages on the oscilloscope trace of the video

waveform obtained from an edge profile are also used


The different sources of noise in digital fluoroscopy

described by Kruger, Mistretta and Riederer and specified in

equation 2.8 were further evaluated and measured by Cohen

et al. (Coh82) by determining the relative importance of

these noise contributors to the total image noise in

conventional fluoroscopy by means of evaluating signal-to-

noise ratios (SNRs) for a particular DSA system. However,

this method requires equipment and conditions atypical of

the clinical setting, which precludes it from routine use in

image quality assessment. Recently, Arnold and Schiebe

(Arn84) have used a more practical approach by studying the

functional dependence of total image noise and SNR on

relative exposure and frame averaging for a limited region

of interest (ROI) in DSA images.

2.7 Low-Contrast Detectability

Subject contrast can be defined as the difference in

the x-ray attenuation property (as a function of x-ray

energy) of a certain object with respect to that of its

surrounding (Wag91). This difference is transformed by the

imaging system into a difference in the image signals

corresponding to the object and its surrounding (e.g.,

optical density on a film, gray shade in a fluoroscopic

image, or pixel value in a digital image) that is known as

image contrast (Kru84). Assessment of low-contrast

detectability is thus a measurement of the ability of a

system to display small image contrast differences.

The low-contrast detectability of a fluoroscopic system

may be described in terms of threshold contrast values; this

quantity is defined as the minimum subject contrast that the

system is capable of displaying for positive visual

detection of an object of a particular size against its

background. Threshold contrast values are not only

dependent on x-ray beam energy, but they are dependent on

three additional parameters (Ros37, Har86, Oha86): object

size, pixel size and signal-to-noise ratio. The

relationship among these quantities is commonly known as the

Rose model of detection.


Harrison and Kotre (Har86) indicated quantitatively the

expected reciprocal relation between threshold contrast and

SNR: lower threshold contrast values result from higher

SNRs. Because, as explained in Section 2.6, quantum noise

represents the major contribution to image noise in

fluoroscopy, higher SNRs result from increased fluoroscopic

exposure rates (Kru84), and therefore, lower threshold

contrast values result from increased fluoroscopic rates.

The application of the Rose model to digital imaging is

explained in detail in Chapter 4.

Routine determination of low-contrast detectability has

historically consisted of imaging specific low-contrast

tools such as penetrameters or specially designed phantoms

such as the Leeds Test Objects (LTO) (Cow87) and using the

visual perception of one or more observers to determine the

number of low-contrast objects that can be visually detected

from the image of such a phantom. These procedures,

although easy to perform, are extremely subjective and time-

consuming for the results thereby obtained.

Newer and better digital imaging systems have made the

ability of a system to display low-contrast objects in a

noisy background a crucial parameter in the assessment of

image quality for the purposes of monitoring performance and

equipment selection (Hen85). Initial work in the evaluation

of low-contrast detectability in nuclear medicine by Myers

(Mye85) was redeveloped and adapted for digital fluoroscopy


by Harrison and Kotre (Har86). In their work, Harrison and

Kotre measured signal-to-noise ratios for a specific ROI as

a function of exposure rate, calculated threshold contrast

values and obtained contrast-detail curves for a particular

digital fluoroscopy system.

Contrast-detail curves were originally developed by

Cohen (Coh85) to show graphically the relationship among

object size, threshold contrast value and radiation dose as

expressed in the Rose model. Ohara et al. (Oha86), using a

method based on statistical decision theory, predicted and

measured low-contrast detectability for simulated low-

contrast objects.

To date, however, there is not a widely accepted method

for low-contrast detectability in fluoroscopy that provides

reliable, objective information; as a consequence, no

accepted performance standard currently exists for the

assessment of low-contrast detectability in fluoroscopy.

2.8 The Use of Carbon Dioxide in Digital
Subtraction Angiography

For years, angiographic studies have made use of

contrast agents and the subtraction of radiological images

in cardiovascular procedures for the observation of blood

vessels (Cru80). However, it was not until the late 1970s

that research groups at the University of Arizona and the

Children's Clinic at Kiel, Germany, first employed digital

techniques for subtraction angiography procedures (Kru84).

The availability of digital subtraction has produced

tremendous advances in angiographic procedures, which have

become an essential part of present-day radiology


Various chemical agents have been introduced into blood

vessels and other anatomical structures for the purpose of

enhancing the contrast of such anatomical structures

relative to an adjacent one. Reasons for the selection of a

particular substance, or contrast agent, depend not only on

radiological but also on biological criteria, such as

toxicity and clearance rates, and physical properties, such

as viscosity, cohesion and adhesion characteristics. During

the first years of clinical use of DSA, iodine-based

compounds were used both intravenously and intraarterially,

as they had been in common use in traditional subtraction

techniques. The new digital techniques reduced

substantially the amount of injected contrast required in

intraarterial procedures for acceptable image quality;

however, that was not the case for intravenous use, and such

procedures have been largely abandoned in the United States

because of unsatisfactory image quality as compared with

intraarterial studies (Ver91, Wea91).

Clinical and imaging advances in vascular surgery,

along with innovations in modern anesthesia and critical

care, have made interventional procedures more feasible for

patients with histories of renal, cardiac or other medical

illnesses. However, the toxicity and allergic potential of

iodine-based intravascular contrast agents remains

significant, and thus for many patients the performance of

angiographic studies still represents a hazard.

The use of gases as contrast agents in radiological

procedures dates back to the early decades of the century

(Alv21). Air and oxygen were the obvious first choices in

this matter; however, after a high number of cases of severe

gas embolism were reported in the 1920s, their use was

discouraged until the 1950s, when Burko (Bur51) reported

excellent results with carbon dioxide (C02) as an intra- and

retroperitoneal, as well as an intravascular, contrast


Use of carbon dioxide: advantages. In more recent

years, Hawkins (Haw82, Haw85) has developed protocols and

injection instrumentation for the use of CO2 as an

intraarterial contrast agent, reporting high-quality

arteriographic images of the abdominal aorta, visceral and

renal arteries and of the lower extremity arterial tree.

Carbon dioxide has several characteristics that make it an

important alternative to iodinated agents. Some of these

characteristics can be listed as follows:

Carbon dioxide is nonallergenic; therefore, it

eliminates any risk of fatal hypersensitivity reactions

(Haw82, Haw85).

Because of its nontoxicity, CO2 eliminates the need

for prearteriography hydration in patients with coexistent

cardiac and renal dysfunction (Wea91).

It is possible to use carbon dioxide in sequential

studies on consecutive days without increasing the risk of

renal failure (Haw82, Wea91).

Carbon dioxide is extremely inexpensive: presently,

a tank of CO2 gas has an approximate cost of 20 dollars,

compared to commonly-used iodinated contrast agents, which

typically cost about 1 dollar per milliliter (Haw85, Wea91).

Use of carbon dioxide: disadvantages. Some of the

difficulties encountered in the clinical application of this

new methodology include the compression of the bolus during

the first seconds of intraarterial administration of CO2,

segmentation of the bolus upon delivery and the natural

buoyancy of the gas (Haw85). The problem of delivery of the

bolus becomes more severe when CO2 is injected by hand,

since an explosive or segmented delivery can occur very

easily (Wea91). In order to solve this problem, an

automated injection device for CO2 very much like that

currently used for iodine-containing contrast agents has

been developed for clinical use in the Department of

Radiology in Shands Hospital at the University of Florida.

Another major difficulty in imaging with this new

methodology arises from the decreased lower subject contrast

of C02, which is a consequence of its lower x-ray


attenuation properties, as compared with those of iodinated

contrast media. This lower inherent signal contrast and the

constraints on image quality produced by image noise in the

carbon dioxide DSA images affect the clinical utility of CO2

in DSA. For this reason, an in-depth study on the

optimization of the low-contrast detectability of gas-filled

vessels (or organs in general) on subtracted and

unsubtracted DSA images is necessary. To achieve this main

objective, specific low-contrast detectability studies of

the dependence of threshold contrast on x-ray beam quality

and levels of radiation exposure were performed for

optimization of image quality for intraarterial use of CO2

in DSA. Chapter 6 presents the methodology developed for

this purpose, and the results obtained are presented in

Chapter 7.


Since the development of digital subtraction

angiography (DSA) in the late 1970s, digital capabilities

for fluoroscopic images have been restricted to dedicated

DSA or digital fluoroscopy systems, i.e., those that include

a mainframe computer that controls and performs the tasks of

digitization, image operations and display of the resultant

digitized fluoroscopic images. Typically, only video frames

coming from fluoroscopic systems which are physically

hardwired to a mainframe computer can be digitized and

processed, as in the case of dedicated digital systems. It

is not possible to process any other video signal from

another conventional or digital fluoroscopic unit when no

interface exists between such a system and the computer.

An alternative to hardwired digital systems is provided

with the use of frame grabbers, which allow the digitization

of analog images from conventional fluoroscopy systems.

These are devices capable of capturing a continuous analog

video frame and converting the various video levels in that

single frame into an array of integer numbers, i.e., pixel

values, to form a digital image. A particular pixel value

is represented by a shade of gray in the displayed digital

image. The appearance of the digitized image and its

similarity to the displayed analog image depends on various

factors, such as sampling, maximum number of gray shades

represented, display matrix, and others. These factors are

discussed later in this chapter.

In spite of the potential advantages from digitization

of an analog fluoroscopic signal, a frame grabber connected

to a conventional fluoroscopic system does not introduce by

itself any clinical advantage when no other hardware for

digital image handling and processing is present, and

therefore it is not common to find such a device permanently

interfaced to the video system of a conventional fluoroscopy

unit in order to provide digitized analog images which are

readily available at all times.

3.1 The FluoroVision(TM) Image Enhancement System

The possibility of having a frame grabber connected to

a conventional fluoroscopic unit as a component of a simple

digital image processing device of known clinical utility

was recently achieved, when Schmidt et al. (Sch88) reported

the development of a portable, real-time modular image

processor specifically designed for clinical use in

conventional fluoroscopy. Such a device has been

manufactured by DigiVision, Inc. with the commercial name


"FluoroVision(TM) Image Enhancement System" (1). Because of

its portable characteristics, the FluoroVision Image

Enhancement System (2) can be connected to a conventional

fluoroscopic system on either a temporary or on a permanent


The FluoroVision is primarily designed to perform noise

reduction and image enhancement operations on analog

fluoroscopic images. Noise reduction is carried out through

simple frame averaging; image enhancement is performed using

algorithms for background removal and edge and contrast

enhancement (Sch88). No special hardwired connection for

installation of the FluoroVision is needed; it is simply

installed by direct connection to the video output of the

conventional fluoroscopic unit. A description of the main

characteristics of the FluoroVision is presented in the

following sections.

3.1.1 Frame Grabbing and Digitization

The frame grabbing and digitization unit of the

FluoroVision is responsible for the following tasks:

a. System timing and control. This part of the

FluoroVision is in charge of video-related control signals

such as horizontal and vertical synchronization, blanking

(1) FluoroVision is a registered mark of DigiVision Inc.
(San Diego, CA) and Medrad Inc. (Pittsburgh, PA).

(2) The FluoroVision Image Enhancement System is referred
henceforth simply as "the FluoroVision".

and wideband video, with the video signal coming from the

fluoroscopy unit.

b. Image sampling. The FluoroVision digitizes input

video onto a nominal 512x512 array. The unit is compatible

with RS-170 standard video, i.e., 525 lines, 60 fields per

second, which, in the case of fluoroscopy, results in having

the entire circular field of the image intensifier being

contained in a 480x480 array. This reduction of the

effective image matrix is expected and results from two main

reasons: first, the actual number of active video lines and

second, the aspect ratio of 4:3 (Ver91) for 525-line

television. In other words, even though a 512x512 memory is

incorporated in the FluoroVision, only 480x480 is

effectively used to store and display the inscribed circle

of the fluoroscopic image.

c. Signal digitization. Analog video levels are

digitized into pixel values. The FluoroVision converts the

signal into an array 8 bits deep, i.e., 256 grey levels,

with zero corresponding to inactive video.

3.1.2 Image Enhancement and Noise Reduction

The main unit of the FluoroVision is responsible for

the tasks of noise reduction and image enhancement

operations. A brief description of these image operations


a. The image enhancement unit utilizes a modified

Wallis algorithm (Pra78) to map pixel values of the

digitized input image (PVinput) onto the enhanced output image

(PVutput). The algorithm makes use of the following


PVoutput = C(PVinput PVt) + LMd + (1 L)nput (3.1)

where the first term produces a zero-mean difference image

with a gain (C) which can be changed without exceeding the

dynamic range of the display monitor. The second and third

terms add a user-defined portion (L) of the desired mean

average brightness of the output (Md). These three

parameters, C, L, Md are user-defined, and the user has

control over them at all times with the use of three

corresponding operator-controlled knobs. The algorithm

reduces veiling glare and lowers brighter areas closer to

the desired mean average brightness.

b. Noise reduction is performed in the frame averaging

and storage unit, in which different levels of averaging can

be performed. For both normal fluoroscopy and cine, 2, 4 or

8 frames can be averaged for noise reduction. The

particular FluoroVision unit used in this work also provides

the capability of storage of up to 100 frames (or averaged

frames) into hard drive memory.

Digital processing and frame averaging as provided by

the FluoroVision are indeed valuable features in the

clinical environment. However, to evaluate a fluoroscopic

system properly, it is extremely important to analyze raw,

i.e., unprocessed, unaveraged test images. Unprocessed

images may be obtained first, by switching off the averaging

function of the FluoroVision and, second, by setting C to 1

and L to 0; these values convert equation 3.1 to PVtput =

PVinput, i.e., the image output of the FluoroVision is an

unprocessed, unaveraged image.

3.2 Communications Interface

Access to the digitized fluoroscopic images is possible

through the RS-232 communications port of the FluoroVision,

which permits pixel value transmission back and forth for

data handling and image post-processing. The RS-232

communications port and the portable nature of the

FluoroVision provide the experimental hardware which has

made possible the realization of this work, for they have

created the possibility of obtaining objective data, i.e,

unprocessed pixel values, from images obtained from

fluoroscopic systems with standard RS-170 video.

Communication between the FluoroVision and an IBM-

compatible laptop personal computer has been achieved

through a direct interface and use of the commercial

communications software package Bitcom 3(C) (Everex, Inc.,

Fremont, CA). The FluoroVision uses two different formats

for representation of pixel values for data transmission or

display: binary format and hexadecimal format. Because

hexadecimal string representation of pixel values in the 0

to 255 range makes storage and string handling simpler than

binary, the hexadecimal format was selected for data

acquisition in this project.

The FluoroVision has a specific command for

transmission of pixel values. This command requires

specification of a region of interest (ROI), which is an

advantage when only limited portions of the test images are

of interest; however, when the entire image is needed, it

cannot be transmitted in one step. Memory restrictions in

the FluoroVision set an upper limit in the ROI size to

blocks of 64x480 pixels within the effective 480x480 matrix.

It is possible for the user to program Bitcom 3(C) with

the proper FluoroVision commands to collect data

automatically from specific ROIs in a test image. These

specific dedicated transmission procedures that

automatically retrieve selected data for the specific tests

and experiments described in Chapter 4 are described in

detail in Appendix A. These programs were designed to be

user-friendly, so that only a few simple commands are

required for their operation.

3.3 Dedicated Software

As it has been previously explained, the nature of this

work required the development of dedicated software to carry

out various tasks and calculations. Examples of the tasks

carried out by the software developed include:

a. Data transmission procedures for different groups

of ROIs of various sizes. As mentioned in Section

3.2, these procedures were written in the

communications language of Bitcom 3(C). Specific

descriptions of these protocols are included in

Appendix A.

b. A routine for conversion of hexadecimal data

strings into pixel values. This subroutine is

called CONV, and it appears in the first stages of

the programs mentioned below. It is described in

detail in Appendix B.

c. Image averaging algorithms. The subroutine AVER,

also described in Appendix B, carries out a

pixel-by-pixel averaging of images for specific

ROIs. For all the individual programs for

calculation of image quality parameters, AVER is

called after CONV has been completed.

d. Calculation of image quality parameters. Each of

the tests described in the next chapter makes use

of a dedicated algorithm and program. These are

described in detail in Appendix B.

CONV, AVER and the remainder of the specific programs

use a TURBOBASIC(R) (Borland International, Inc., Scotts

Valley, CA) interpreter for IBM-compatible computers. In

this way, software characteristics match the versatility and

transportability of the equipment used in the project. More

important, the basic goal of simplicity for this work, as

described in Chapter 1, is maintained through simple,

easily-accessible programming language and software.

3.4 Test Objects, Phantoms and Other Equipment

In developing a simple set of test procedures, an

important goal, following the basic premise of simplicity

described in Chapter 1, was to design these procedures so

that test objects and phantoms in common use for the

subjective assessment of image quality in conventional

fluoroscopy could be also utilized for the objective

methods. Thus, traditional phantoms and test objects can be

used with these objective methodologies without having to be

modified or redesigned.

In all of the cases described in the next chapter,

phantoms and test objects were chosen because of their

availability in the Department of Radiology in Shands

Hospital at the University of Florida. Nonetheless, similar

phantoms may be used in most of the cases without having to

modify the algorithms and programs substantially. All

phantoms are described in detail in their corresponding

sections in Chapter 4.

For the case of low-contrast detectability of air-

filled objects in DSA, a basic patient-simulating phantom

consisting of lucite blocks was used together with the

Nuclear Associates (3) DSA phantom, which basically

consists of an abdomen-simulating phantom and several lucite

inserts for assessment of image quality in iodine-DSA.

Thus, it was necessary to design and manufacture additional

lucite inserts for the assessment of low-contrast

detectability of air-filled objects. The phantom and the

corresponding inserts are described in detail in Chapter 6.

Whenever it was necessary to determine and monitor x-

ray tube voltage (kVp), a Keithley Model 35080A Noninvasive

kVp Divider along with a Keithley Model 35050A Digital kVp

Readout were used (4). Appropriate range packs were used

in the kVp divider as required depending on the x-ray tube

voltage selected.

In the case of exposure rate measurements, most

dosimeters available for primary diagnostic x-ray beams can

be conveniently used for the purposes of these test

(3) Nuclear Associates is a division of Victoreen, Inc.
(Carle Place, NY).

(4) Manufactured by Keithley Instruments, Inc.
(Cleveland, OH).

procedures. A Radcal MDH(R) (5) Model 1015 Series X-ray

Monitor (6) was used to perform such measurements in this

work. Entrance exposure rate measurements were performed

using a model 10X5-6 (6 cc) ion chamber, and exposure rates

measured at the image intensifier input phosphor were

performed using a model.10X5-60 (60 cc) ion chamber ().

3.5 Fluoroscopic Systems Used

Four conventional fluoroscopy systems were evaluated

for the purposes of the first portion of this work. For the

purposes of identification, Table 3.1 lists the units, the

nominal dimensions of their image intensifier modes and

their identification used henceforth. Detailed description

and specifications of these systems are included in Appendix

C. With the exception of system D (a Philips(R) Modular

unit), which was taken out of commission in September, 1992,

the systems are in current use in the Department of

Radiology in Shands Hospital at the University of Florida.

As it is further explained in Chapter 6, fluoroscopic

System B was chosen to perform the studies on the low-

contrast detectability of air-filled objects mainly because

(5) Manufactured and distributed by Radcal Co., a division
of MDH Industries, Co. (Monrovia, CA).

(6) The Radcal MDH(R) x-ray monitor is referred henceforth
simply as "the MDH monitor".

(7) Dosimeter and ion chambers were calibrated at the
Radcal Inc. (Monrovia, CA) NIST traceable facility on
December 26, 1991.


of its capability to control tube voltage (kVp) manually

during fluoroscopy.





A brief review of the historical evolution of

qualitative methods for image quality assessment in

conventional fluoroscopy and the procedures typically used

for the subjective determination of image quality parameters

was presented in Chapter 2. Using these methodologies as a

baseline, the main objective of this part of the project was

to develop a complete standardized set of simple procedures

for the objective (quantitative) assessment of image quality

in conventional fluoroscopy, as expressed in Chapter 1. The

present chapter describes, in a step-by-step fashion, the

methods and test procedures developed for the determination

of the image quality parameters described in Chapter 2. As

it was previously indicated, analysis of the digital data

obtained from these test procedures is carried out using

dedicated software specifically written for this project.

The specifics of the algorithms and corresponding BASIC

programs are included in Appendix B, and only the basic

steps of these algorithms are outlined in the corresponding

sections in this chapter.


Figure 4.1 details the general setup for acquisition of

test images. Although some variations exist for specific

test procedures, the modifications are indicated in the

corresponding sections. As indicated in the figure, all

test objects or phantoms are positioned on top of a hollow

plastic support (point A) and centered in the fluoroscopic

field for imaging; points B and C indicate locations at

which exposure rates are measured. Entrance exposure rates

(EER) are measured at point B, while exposure rates measured

at the image intensifier input surface (IER) are measured at

point C. For units with spot film capabilities, IER

measurements are obtained by placing the dosimeter probe

inside the spot film device. The tabletop to image

intensifier input distance, i.e., the distance from the top

of the table to the image intensifier input phosphor plane

at the center of the field, is set to approximate 40 cm.

Unless otherwise specified for a particular test,

fluoroscopic images are obtained under conditions of

automatic brightness control (ABC).

The set of objective test procedures described in the

sections below includes the determination of the following

parameters, as previously listed in Chapters 1 and 2:

brightness uniformity, spatial linearity, system contrast

ratio, relative conversion efficiency, image noise,

modulation transfer function and low-contrast detectability.

Figure 4.1. Test equipment setup. Tabletop-to-image receptor distance = 40 cm.
A: Position for imaging test objects and phantoms. B: Position for EER
measurements. C: Position for IER measurements. Points are located in the
center of the fluoroscopic image.


Each one of these procedures is detailed in the following


4.1 Brightness Uniformity

The research objectives originally proposed for this

section included, first, the development of a method for

objective assessment of brightness uniformity and, second,

the development of correction-factor matrices for

nonuniformities. In this methodology, both integral

uniformity (IU) and a 480x480 matrix of correction factors

can be specifically calculated for each image intensifier

mode of the fluoroscopic unit being evaluated (1). The

following sections detail the procedures for the

quantitative assessment of brightness uniformity.

4.1.1 Test Procedure for Calculation of Integral Uniformity

The methodology developed for assessment of brightness

uniformity by means of a quantitative determination of

integral uniformity is based on the NEMA standard for gamma

cameras (Mue81, Nat86), as described in section 2.1.1. It

has been adapted for the case of fluoroscopy as indicated in

the following subsections.

(1) Actually, a 512x512 correction matrix is calculated;
however, pixel values outside the 480x480 matrix are
always zero, as indicated in Section 3.1.


The phantom for this test consists of a four-inch

thickness of acrylic of sufficient size to cover the entire

nonmagnified fluoroscopic field. Such a setup is referenced

in this document as "the uniformity phantom". Testing procedure

Step 1: Obtaining a test image. The equivalent of the

gamma camera flood-field image is obtained in fluoroscopy by

imaging the uniformity phantom under ABC conditions and

capturing and digitizing its fluoroscopic image. The entire

digital image is transmitted and stored in memory.

Step 2: Smoothing of data. The equivalent of the NEMA

smoothing algorithm for gamma cameras is obtained for the

fluoroscopic image of the uniformity phantom by applying an

averaging algorithm to an adequate number of images. The

dedicated program for the assessment of brightness

uniformity is called UNIF (2). As it was mentioned in

Section 3.2, during the initial stages of a run, UNIF

activates the pixel-by-pixel averaging subroutine called

AVER (3) that can average up to 25 images for the purpose

of reducing random fluctuations.

Figure 4.2 presents an example to demonstrate the

impact of the averaging operation on fluoroscopic images as

obtained in fluoroscopic system D. First, Figure 4.2a shows

(2) See Appendix B for a detailed description.

(3) See Appendix B for a detailed description.

Figure 4.2. Effect of image averaging on random
fluctuations: a. Profile of the central row
of a single, uncorrected image. b. Profile of
the central row for the corresponding averaged,
uncorrected image (obtained using 10 images).



m 80

> 60




0 100 200 300 400 500





w 80

> 60

aC 40



0 100 200 300 400 500




the central row profile obtained from a single, uncorrected

image of the uniformity phantom. Superimposed on the

nonuniform brightness and vignetting of the image are

fluctuations in brightness among adjacent regions. Second,

Figure 4.2b shows the result after 10 images of the phantom

have been averaged; it is clear that random fluctuations

have been greatly reduced. The presence in Figure 4.2b of

the remaining visually apparent fluctuations indicates that

image averaging alone does not produce a totally smooth

profile, because structural nonuniformities remain. The

sources of this type of nonuniformity were previously

discussed in section 2.1.2.

Step 3: Obtaining quantitative data. UNIF identifies

the maximum (PVmax) and minimum (PVin) pixel values in the

averaged 480x480 matrix. For the example shown in Figure

4.2, it is found that PVax = 112 and PVm = 38.

Step 4: Calculation of image quality parameterss.

Integral uniformity is calculated by UNIF using the

equivalent of equation (2.1) as applied to fluoroscopy as


( PVmax PVeg )
IU = 100 PVax PVin (4.1)
( PVax + PVin )


IU = Integral uniformity,

PVmax = Maximum pixel value in the image matrix, and

PVn = Minimum pixel value in the image matrix.

For the example shown in Figure 4.2, IU = 49.3%.

According to equation 4.1, an IU of 0 indicates an image

which is completely uniform.

4.1.2 Nonuniformity Corrections

The typical algorithm for the correction of brightness

nonuniformities in gamma cameras consists of normalizing all

image matrix elements to the element with the maximum number

of counts (Sor87). The equivalent of this factor in the

case of conventional fluoroscopy is PVax. Pixel-by-pixel

correction factors are then calculated by UNIF using

CF j PVax (4.2)
PV ij


CFi, = Correction factor for pixel location i,j, and

PVi, = Original pixel value for pixel location i,j.

Note that CFij 1. The calculation of correction

factors is carried out by the UNIF program at the request of

the user, and a particular run can be terminated after

determination of IU alone. Calculated correction factors

are characteristic of the fluoroscopic unit and the

particular image intensifier magnified mode selected; they

can be saved for correcting nonuniformities caused by the

sources described in section 2.1.2 for stored images

obtained in the particular unit for the specific

magnification mode selected.

4.2 Spatial Linearity

As discussed in Chapter 2, several models and

algorithms have been used over the years for assessment of

spatial linearity and the calculation of correction factors

for nonlinearities in conventional fluoroscopic images. Of

these, the model proposed by Chakraborty (Cha87) was chosen,

because it was specifically developed for digital

fluoroscopic systems, as described in section 2.2, and

because of its simplicity and ease of application. A

description of the original model and its application for

fluoroscopy systems has already been included in Chapter 2.

The following sections present the pertinent

modifications for a simple, quantitative test procedure to

assess spatial linearity in conventional fluoroscopy.

4.2.1 Modifications to the Chakraborty Model for an
Objective Assessment of Spatial Linearity

The research objective contemplated for completion of

this section of the project required the development of a

test procedure to quantify spatial linearity objectively in

conventional fluoroscopy systems. A simple, quantitative

assessment of spatial linearity is obtained by first

combining equations 2.3, 2.4, 2.5 and 2.6 to obtain the


direct relationship between the object coordinates (x,y) and

pixel locations (x",y") in the image matrix as follows:

x/ = M(x,y) Ax x + D, (4.3)

y/ = M(x,y) Ay y + Dy (4.4)

Next, some considerations are needed with respect to

the magnification factor M(x,y), which is indirectly

dependent on the radius of curvature of the image

intensifier input phosphor. This radius is not usually

provided by the manufacturer, and its measurement is very

difficult to perform in the clinical setting (Cha87).

Trying to include such a measurement in the assessment of

spatial linearity would go against the basic premise of

simplicity of this project as discussed in Chapter 1, i.e.,

to develop an uncomplicated set of test procedures for the

assessment of image quality.

It is possible to define the following empirical

quantities for any of the image intensifier magnified modes:

Cx = Ax M(x,y) (4.5)


Cy = Ay M(x,y)


where M(x,y) represents the average value that the

magnification factor can take over the entire useful surface

of the image intensifier input phosphor. Thus, it is

possible to characterize spatial linearity in a simple way

by using equations 4.3 and 4.4 as follows:

x/ = C x + Dx (4.7)

y = Cyy + Dy (4.8)

where Cx, Cy, Dx and D are determined empirically.

4.2.2 Test Procedure for Assessment of Spatial Linearity

The following sections detail the procedure developed

to determine spatial linearity in a conventional fluoroscopy

system quantitatively. Phantom

In the case of spatial linearity, the use of a phantom

containing equidistant lines has proven to be an excellent

choice for subjective methods. The Leeds Test Object (LTO)

M1 (Cow87) phantom was chosen among the ones available for

its easy use and setup. The phantom consists of a set of

perpendicular wires embedded in acrylic. As it is shown in

Figure 4.3, which presents a fluoroscopic image of the LTO

M1 obtained in fluoroscopic system A, the wires in the

phantom are equally separated and placed in perpendicular

Figure 4.3. Fluoroscopic image of the Leeds Test Object Ml
obtained in fluoroscopic system A.


directions. Except for the central row and column, which

have wires 1 cm apart, the separation between wires for the

rest of the phantom is 2 cm. Thus, squares of 2 cm on a

side are available for visualization and evaluation of image

distortion. Testing procedure

Step 1: Obtaining a test image. The LTO M1 must be

carefully positioned on the plastic support in such a way

that the wires are reasonably aligned with the vertical (and

horizontal) directions as observed in the monitor. Because

this alignment is not expected to be extremely precise (40

to 50 misalignment is considered acceptable), this is a

rather easy step. More important to the procedure is the

centering of the object. The program SPCLIN (4) for

assessment of spatial linearity relies on the coordinates

(pixel locations) of the central square in the test object

for calculation of the four spatial linearity parameters Cx,

C DX and D Therefore, it is important to center the test

object visually in the fluoroscopic image to the best

achievable. Figure 4.3 shows an acceptable case of

positioning. The LTO M1 is imaged in the nonmagnified

fluoroscopic field under ABC conditions.

Step 2: Smoothing of data. Smoothing and analysis of

the digital data obtained are performed using the SPCLIN

algorithm. It is important to note, however, that the

See Appendix B for a detailed description.


SPCLIN program requires previous knowledge of the brightness

uniformity correction factor matrix CFi, Therefore, it is

necessary to obtain uniformity data and run the UNIF program

(as specified in sections 4.1.1 and 4.1.2) for the

particular fluoroscopic unit and image intensifier mode for

which spatial linearity is being assessed prior to the use

of the SPCLIN program.

Ten images of the LTO M1 are obtained as indicated

above. The same averaging subroutine employed in the

program UNIF appears in the SPCLIN algorithm for the purpose

of reducing random fluctuations. A typical profile from the

central row of the resultant image, i.e., an averaged,

uncorrected image is shown in Figure 4.4 for a test image

obtained in fluoroscopic system A.

Step 3: Correction of data for nonuniformities.

After averaging of the images has been performed by the

SPCLIN algorithm, the next step consists of accessing the

appropriate CFij correction factor matrix (calculated by the

UNIF program) to correct the averaged test image for

nonuniformities. Figure 4.5 shows the result of this

operation for the case of the central row profile of Figure

4.4. As indicated above, the number of wires appearing in

the central row profile is twice as many as those in other

rows, because of the construction of the test object.

Step 4: Obtaining quantitative data. The next task

for the SPCLIN algorithm involves determination of the

Figure 4.4. Spatial linearity evaluation: effect of image
averaging. Shown is an averaged, uncorrected
profile of the central row.





5 90

> 80-
L 70



0 50 100 150 200 250 300 350 400 450 500

Figure 4.5. Spatial linearity assessment: effect of
applying corrections for image nonuniformities:
shown is corrected row profile corresponding to
the spatial linearity test image in Figure 4.4.

100 150 200 250 300 350 400 450 500






x 70





coordinates (pixel locations) of the wire crossings in the

test image. The actual pixel locations of the wires are

defined to be those corresponding to the local minimum

values (troughs) in a row profile, such as the ones shown in

Figure 4.5. It is at this point that the importance of

using the correction factors CFi, on the original averaged

spatial linearity data is apparent. As it can be observed

in the various troughs shown in the row profiles of Figures

4.4 and 4.5, after the image has been corrected for image

nonuniformities, all the troughs representing the wires have

depths in the range of 72 to 80% of the overall maximum

pixel value.

Because image contrast is dependent on x-ray energy, as

mentioned below in section 4.7.2, the depth of the troughs

corresponding to the wires in the test image of LTO M1 is

dependent on the x-ray beam quality used in imaging the test

object. Thus, imaging the object under ABC conditions in

various fluoroscopic units results in the use of different

x-ray beam spectra for imaging purposes and thus, different

trough depths in test images from different fluoroscopic

units. This potential analytic difficulty is overcome by

the SPCLIN algorithm by finding the overall maximum and

minimum pixel values from the averaged, corrected test image

of LTO Ml, in the same fashion as UNIF performs the task for

the image uniformity case. SPCLIN then recognizes the pixel


value at a certain location PVi, as that corresponding to a

trough if the pixel value satisfies the condition

PVmin : PVi,j < 0.8 PVmax (4.9)

Upon completion of this search, the SPCLIN program

identifies a complete set of (x",y") coordinates of the wire

crossings in the digital test image of LTO Ml.

Step 5: Calculation of image quality parameters. As

specified in section 4.2.1, determination of the spatial

linearity empirical parameters requires knowledge of both

the digital image (x",y") and object (x,y) coordinates of

the wire crossings in the LTO Ml. Because of the fact that

objects in the vicinity of the center of the fluoroscopic

image are essentially undistorted (Cso85), the set of object

coordinates (x,y) can be constructed from the (x",y")

positions of the wire crossings which form the central

square in LTO Ml. Since this square is undistorted, the

"expected", i.e., undistorted positions of all the other

wire crossings in the object can be calculated from it, thus

producing the necessary set of object coordinates (x,y).

Knowledge of both sets of coordinates allows SPCLIN to

determine the coefficients and intercepts of the modified

Chakraborty linear model through a linear regression on

these data. Knowledge of both the expected and actual pixel

locations of the wires provides a way to correct the image

of the test object for distortions. By using the ratio of

expected to actual values, SPCLIN corrects the test image

for distortions. Results of this correction operation are

presented in section 5.2.2.

4.3 System Contrast Ratio

The main objective for this section of the project was

to develop a noninvasive, reproducible, objective test

procedure for determination of contrast ratio. As described

in section 2.3, contrast ratio is a performance parameter

that measures the inherent loss of contrast that occurs in

image intensifier tubes (Bro82). As in the case of the NEMA

standard described in section 2.3.2 (Nat92), the test

procedure described in the present section determines a

value of contrast ratio that includes contributions to loss

of contrast due to the remaining components in the imaging

chain, such as the optical distributor, other optical

devices, the TV system, and, in the case of digital systems,

the analog-to-digital converter.

Strictly speaking, this procedure yields a relative,

rather than an absolute, contrast ratio as defined for image

intensifier tubes and described in section 2.3.1. As

specified in the present section, the values obtained from

performing such a relative determination of contrast ratio

are not comparable with contrast ratio values given in the

image intensifier manufacturer's specifications; however,

they can be used for reliable long-term monitoring of

fluoroscopic units and comparisons among such units.

4.3.1 Definition of System Contrast Ratio.

As described in section 2.3.2, the concept of system

contrast ratio is an expansion of the definition of contrast

ratio for image intensifier tubes to include contributions

to the overall loss of contrast from all components in the

imaging chain. Although this concept is similar to that of

the NEMA standard for determination of system contrast ratio

as described in section 2.3.2 (Nat92), the method for the

determination of the system contrast ratio described in the

present section has been developed keeping the same basic

structure of traditional measurements. Thus, the method

described here does not follow the NEMA standard completely,

because of the limited applicability of the standard as

discussed in section 2.3.2.

As in the case of traditional methods, three different

test images are required for determination of the system

contrast ratio:

Image "BKG" simply represents an averaged image

captured and digitized while video is inactive in the

fluoroscopic unit. Such an image provides knowledge of the

baseline pixel values for the unit.

Image "D" requires an appropriate contrast ratio

object, as defined in section 2.3.2, to be positioned at the

center of the fluoroscopic field.

Image "ND" is obtained after the contrast ratio

object is removed from the fluoroscopic field, but using the

same fluoroscopic techniques (kVp and mA) utilized by the

fluoroscopic unit in obtaining image "D".

By selecting a circular region of interest (ROI)

centered in the fluoroscopic field and calculating the mean

pixel value (MPV) from this ROI in the three test images

described above, the system contrast ratio (SCR) is defined


SCR = PVN MPVB (4.10)


MPVN = Mean pixel value from circular ROI, image ND;

MPVD = Mean pixel value from circular ROI, image D and

MPVB = Mean pixel value from circular ROI, image BKG.

The following section describes the application of this

definition for practical determination of the system

contrast ratio in the clinical setting.

4.3.2 Test Procedure for Determination of the System
Contrast Ratio

In determining the system contrast ratio for a

fluoroscopic unit, the images obtained with and without a

corresponding contrast ratio object centered in the

fluoroscopic field are digitized and analyzed by the

algorithm and program called RATIO (5). The detailed

procedure is described in the following sections. Phantom

As in the case of traditional methods and the NEMA

standard, the objective procedure for determination of the

system contrast ratio makes use of contrast ratio objects as

defined in section 2.3.2 (Nat92). Testing procedure

Step 1: Obtaining test images. As indicated in

section 4.3.1, three different averaged images are required

for determination of the system contrast ratio:

1. Image "BKG" is obtained from capturing frames with

fluoroscopic video inactive.

2. Image "D" is obtained by positioning the

appropriate contrast ratio object at the center of the

fluoroscopic field (preferably in the nonmagnified mode).

The ABC system must be enabled, such that x-ray tube

potential and current (kVp and mA) are automatically

determined by the system.

(5) See Appendix B for a detailed description.

3. Image "ND" is obtained after the contrast ratio

object is removed from the fluoroscopic field. However, the

same kVp and mA values that the ABC system used in obtaining

image "D" are manually set. In order to achieve this, it is

necessary to switch off the ABC system. In the case where

the fluoroscopic unit does not have manual-technique

capabilities, such as in the case of fluoroscopic system D,

the image is obtained under ABC conditions.

Step 2: Data smoothing and collection. Since the

method for determination of the system contrast ratio

described in this section has been developed trying to keep

the same basic structure of traditional measurements, only

data from a 160x160 square ROI centered in the image matrix

are of interest. Such numbers are automatically retrieved

by a dedicated protocol routine called CIRCLE (6) written

in the data transmission software.

For each of the imaging cases described in step 1, ten

frames are obtained; only those pixel values within the

preselected 160x160 ROI are then stored.

Step 3: Obtaining quantitative data. Data from the

square ROI are fed into the RATIO program, which obtains an

averaged image of the ROI and then selects data from a

circle 120 pixels in diameter for all three imaging cases

described in step 1. Such a circular region corresponds to

the shadowed portion of the fluoroscopic field on image "D".

The protocol is described in Appendix A.

RATIO then calculates the mean pixel value (MPV) from the

circular ROI for all three averaged images.

Step 4: Calculating image quality parameters. With

the data obtained, RATIO calculates the system contrast

ratio (SCR) using equation 4.10. It is important to note

that the value of the SCR obtained for a fluoroscopic unit

following the objective procedure given here does not

correspond to either the nominal contrast ratio value

reported in the manufacturer's specifications for the image

intensifier tube alone or to the SCR value obtained using

the NEMA standard. In the first case, the discrepancy

arises from the fact that the system contrast ratio includes

contributions from all the elements in the imaging chain.

In the second case the discrepancy arises from the different

definitions of SCR in the NEMA standard and in the objective


4.4 Relative Conversion Efficiency

In developing a definition for relative conversion

efficiency, a strong effort was made to try to design a test

procedure much simpler than the one described in section 2.4

for determination of conversion efficiency following the

recommendations in ICRU Report 10f (Int62, Nat79), but which

would still yield valuable information about the x-ray to

light conversion capabilities of the fluoroscopic system.

Historically, and in close analogy to the case of

contrast ratio, the conversion efficiency parameter

describes the ability of an image intensifier tube to

convert an x-ray beam of standard characteristics and

intensity into light intensity at the image intensifier

output phosphor. A relative measurement of conversion

efficiency, as described in the present section, measures

the overall conversion capabilities of the complete imaging

chain instead, by looking at the resultant average pixel

value. Thus, a measurement of the relative conversion

efficiency includes the inherent variations of the video

signal intensity throughout the entire imaging chain.

The basic structure of the algorithm RATIO was used

and modified to obtain a simple method of measuring relative

conversion efficiency. The corresponding dedicated program

CONVEFF (7) takes data from a circular ROI in an averaged

blank image for calculation of the relative conversion

efficiency. The following sections describe the specific

steps to carry out this determination in the clinical


4.4.1 Definition of Relative Conversion Efficiency.

Although common methodologies for the measurement of

conversion efficiency, as those described in Chapter 2,

establish a procedure that requires measurement of the image

(7) See Appendix B for a detailed description.

intensifier output luminance or output video signal level

resulting from a well-defined, standard x-ray beam and input

phosphor exposure rate (IER), it is also possible to use a

"reverse" approach to this test and measure the IER

necessary to produce an specific output illuminance or

output video level (Hen85).

The procedure for objective determination of the

relative conversion efficiency described here makes use of

the original approach by measuring the average pixel value

that results from obtaining a blank averaged image similar

to the image "ND" described in section 4.3.1, using a

specific x-ray beam, as well as its corresponding IER. A

circular ROI is selected from this single averaged image; by

calculating the MPV from this ROI, the relative conversion

efficiency (RCE) is defined by

RCE = (4.11)


RCE = Relative conversion efficiency, in sec/mR;

MPVN = Mean pixel value from circular ROI in image ND and

IER = Image intensifier input exposure rate, in mR/sec.

The application of this definition in the clinical

environment is outlined in the following section.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs