Abstract
Contents
Acknowledgements i
Abstract Mi
Preface xi
Contributions xiii
1 Introduction 1
1.1 Breast Cancer 1
1.1.1 Incidence and mortality 1
1.1.2 Staging and prognosis 1
1.2 Breast Anatomy 3
1.2.1 Anatomical structure of breasts 3
1.2.2 Glandular tissue and breast density 5
1.3 Breast Imaging Techniques for Screening and Diagnosis 5
1.3.1 Mammography 6
1.3.2 Other image techniques utilizing ionizing radiation 8
1.3.3 Magnetic resonance imaging and ultrasonography 8
1.3.4 Spiral breast CT equipped with photon-counting detector
technology 9
1.4 Imaging Radiation Dose and Secondary Cancer Risk 10
1.4.1 Effect of ionizing radiation 10
1.4.2 Radiation dose assessment 11
1.4.3 Mean glandular dose 12
1.5 Aim and Outline of the Thesis 12
1.5.1 Requirements for BCT 12
1.5.2 Aim of the thesis 13
2 Methods 17
2.1 BCT with Photon-Counting Detector Technology 17
2.2 Monte Carlo Radiation Dose Simulation 20
3 Lesion Detectability and Radiation Dose in Spiral Breast CT with
Photon-Counting Detector Technology: A Phantom Study 23
3.1 Contributions 24
3.2 Abstract 25
3.3 Introduction 26
3.4 Materials and Methods 27
3.4.1 Image acquisition 27
3.4.2 Image reconstruction 28
3.4.3 Dedicated breast phantom 28
3.4.4 Objective image quality: signal-to-noise ratio and contrast-to-
noise ratio assessment 29
3.4.5 Subjective image quality: visual inspection assessing
detectability 30
3.4.6 Validation of Monte Carlo simulation 31
3.4.7 Dose assessment 32
3.4.8 Exposure setting optimization 32
3.5 Results 32
3.5.1 Signal-to-noise-ratio 32
3.5.2 Lesion detectability 33
3.5.3 Validation of Monte Carlo based on the ionizing chamber 34
3.5.4 Radiation dose 35
3.5.5 Contrast-to-noise ratio as a function of average absorbed dose
3.5.6 Radiation exposure setting optimization 35
3.6 Discussion and Conclusions 36
3.7 Acknowledgements 36
3.8 Appendices 39
40
4 Spiral Breast Computed Tomography (CT): Signal-to-Noise and Dose
Optimization Using 3D-Printed Phantoms
4.1 Contributions 43
4.2 Abstract 44
4.3 Introduction 45
4.4 Materials and Methods 46
4.4.1 Datasets 47
4.4.2 3D printed breast models 47
4.4.3 Tissue phantom (filling) 47
4.4.4 Phantom scans 48
4.4.5 Evaluation 49
4.4.6 Evaluation of average dose by Monte Carlosimulation 50
4.4.7 Validation of the Monte Carlo simulation bydose measurements 51
using MOSFET sensors
4.5 Results 52
4.5.1 Subjective image quality evaluation 52
4.5.2 Objective image quality evaluation 52
4.5.3 Validation of the Monte Carlo calculations:surface dose 52
comparison with MOSFET
4.5.4 Dose distribution 55
4.5.5 Average absorbed dose as a function of breast size and
glandularity 57
4.5.6 Optimization of acquisition parameters regarding image quality
and dose 58
4.6 Discussion 59
4.7 Conclusion 62
4.8 Acknowledgements 62
4.9 Appendices 62
5 Diagnostic Values of Spiral Photon-Counter Detector Breast CT in
Patients with Breast Implants - Initial Experiences 65
5.1 Contributions 66
5.2 Abstract 67
5.3 Introduction 68
5.4 Methods 69
5.4.1 Breast CT 70
5.4.2 Reading and statistical evolution 70
5.4.3 Dose calculation 71
5.5 Results 71
5.5.1 Patient characteristics 71
5.5.2 Breast density 71
5.5.3 Breast Implants 72
5.5.4 Breast lesions 74
5.5.5 Intermodality comparison BCT vs. MRI 75
5.5.6 Average dose 77
5.6 Discussion and Conclusions 78
5.7 Acknowledgments 80
6 Fully Automated Breast Segmentation on Spiral Breast CT images 81
6.1 Contributions 82
6.2 Abstract 83
6.3 Introduction 84
6.4 Methods 87
6.4.1 3D breast image by a novel breast CT with photon-counting
detector technology 87
6.4.2 Dataset 87
6.4.3 Preliminary study on the HU values of the breast components in
the BCT images 88
6.4.3.1 Reference manual segmentation by radiologists 88
6.4.3.2 Histogram processing on HU values of the adipose
and glandular tissues 89
6.4.3.3 Standard HU measurement for each component 90
6.4.4 Fully automatic volumetric segmentation and breast analysis
method 91
6.4.4.1 Investigation of presence and seed region of
components 92
6.4.4.2 Segmentation of components except the adipose and
glandular tissues 94
6.4.43 Quantitatively estimated individual breast density 96
6.4.4.4 Adipose and glandular tissues’ segmentation and
volumetric breast density estimation 96
6.4.5 Segmentation evaluation 96
6.4.5.1 Objective evolutions of automatic segmentation
quality and manual segmentation reproducibility 96
6.4.5.2 Subjective segmentation quality evaluation: Likert
scale 98
6.4.6 Breast density estimation error assessment 98
6.5 Results 99
6.5.1 Evaluation of reproducibility for manual segmentation in BCT
images 99
6.5.2 Evaluation of breast component’s presence investigation 100
6.5.3 Evaluation of strongly absorbing component segmentation 101
6.5.4 Evaluation of soft tissue components segmentation by adaptive
seeded watershed algorithm 101
6.5.5 Evaluation of the glandular tissue segmentation: adaptive region
growing algorithm 102
6.5.6 Quantitative breast density estimation 103
6.5.7 Processing time of the computation and manual segmentation
for BCT images 104
6.6 Discussion and Conclusions 104
7 Radiation Dose Estimates Based on Monte Carlo Simulation for Spiral
Breast Computed Tomography Imaging in a Large Cohort of Patients 109
7.1 Contributions 110
7.2 Abstract 111
7.3 Introduction 113
7.4 Material and Methods 114
7.4.1 Study participants 114
7.4.2 Image data 114
7.4.3 Image segmentation 115
7.4.4 Determination of breast morphologicalfeatures 115
7.4.5 Monte Carlo simulation 116
7.4.6 Determination of radiation dose values 117
7.4.7 Comparison to other X-ray breast imaging methods 117
7.4.8 Statistical analysis 117
7.5 Results 118
7.5.1 Study participants and morphological breast characteristics 118
7.5.2 Segmentation and Monte Carlo simulationresults 119
7.5.3 Morphological features and radiation dosevalues 120
7.5.4 Correlation analysis 122
7.5.5 Statistical analysis on radiation dose 122
7.6 Discussion 125
7.7 Conclusions 131
7.8 Supplementary Material for Analysis and Comparison of Prediction
Models 131
7.8.1 Residual plots of the mono- and bi-exponential regression
models 131
7.8.2 Multi-linear regressor models for dose values estimation 133
7.8.3 Support vector regressor models for dosevalue prediction 135
7.9 Conflict of Interest 136
8 Quantitative Analysis of Breast Composition in a Large Cohort of
Screening Patients Undergoing Photon-Counting Breast CT 137
8.1 Abstract 138
8.2 Introduction 139
8.3 Material and Methods 141
8.3.1 Study participants 141
8.3.2 Image data 141
8.3.3 Image segmentation and component localization 142
8.3.4 Breast composition feature acquisition 142
8.3.5 Breast composition feature analysis 143
8.3.6 Statistical analysis 143
8.4 Results 143
8.4.1 Segmentation and localization results 143
8.4.2 Study participants and average composition feature 144
8.4.3 Composition feature analysis across the age groups 144
8.4.4 Feature analysis in each quadrant 147
8.4.5 Feature analysis in each quadrant across theage groups 148
8.5 Discussion 148
8.6 Appendices 152
9 Discussion and Conclusions 155
9.1 Assessment of Diagnostic Performance of BCT 156
9.2 Development of an Automatic Breast Segmentation Method 157
9.3 Characteristics of Breast Morphology 158
9.4 Radiation Dose Level of BCT 158
9.5 Standard Radiation Dose Level with Optimal Exposure Setup 159
9.6 Prediction of Radiation Dose Values 159
9.7 Quantification of Breast Composition: Relation to Breast Cancer Risk 160
9.8 Outlook and Conclusion 161
Bibliography 163
Sojin, Shim