Vascular and Intravalcular Imaging Trends, Analysis, and Challenges: Plaque Characterization

PRELIMS.pdf
Preface
Editor biographies
Petia Radeva
Jasjit S Suri
List of contributors
CH001.pdf
Chapter 1 Coronary and carotid artery calcium detection, its quantification and grayscale morphology-based risk stratification in multimodality big data: a review
1.1 Introduction
1.2 Calcium detection in coronary and carotid arteries
1.2.1 Calcium detection in coronary arteries
1.2.2 Calcium detection in carotid arteries
1.3 Calcium area/volume quantification in coronary and carotid arteries
1.3.1 Calcium area/volume quantification in coronary arteries
1.3.2 Calcium area/volume quantification in carotid arteries
1.4 Metrics for performance evaluation for calcium detection algorithms and its validation
1.4.1 Statistical metrics for performance evaluation
1.4.2 Validation of calcium detection algorithms
1.5 Machine-learning-based risk stratification
1.5.1 Coronary risk assessment using ML-based approaches
1.5.2 Carotid risk assessment using ML-based approaches
1.6 Discussion
1.6.1 A note on the usage of calcium detection techniques in coronary and carotid arteries
1.6.2 A note on the usage of calcium quantification techniques in coronary and carotid arteries
1.6.3 A note on the use of statistical metrics for the evaluation of calcium detection algorithms
1.6.4 A note on feature selection in ML-based risk stratification for the coronary and carotid arteries
1.6.5 Recommended interventions for CVD patients
1.6.6 Atherosclerotic calcium in coronary and carotid imaging: ongoing challenges
1.7 Conclusions
Conflict of interest
Acknowledgements
Funding
References
CH002.pdf
Chapter 2 Risk of coronary artery disease: genetics and external factors
2.1 Introduction
2.2 External factors
2.2.1 Ethnicity and CVD
2.2.2 Environmental factors and CVD
2.2.3 Air pollution and CVD risk
2.2.4 Nutrition and CVD risk
2.2.5 Family history and CVD risk
2.3 Genetics of coronary artery disease
2.3.1 Genetics of atherosclerosis
2.3.2 Genetics of diabetes
2.3.3 Genetics of rheumatoid arthritis
2.3.4 Anatomy of a 3D heart
2.4 Multimodal coronary imaging
2.4.1 Regular coronary artery
2.4.2 Coronary imaging using x-ray angiography
2.4.3 Coronary imaging using magnetic resonance angiography
2.4.4 Imaging coronary CT angiography
2.4.5 Coronary artery interpretation by OCT versus IVUS
2.5 Association of CVD with other prevalent diseases
2.5.1 Relationship between coronary artery and carotid disease
2.5.2 The relationship between diabetes and coronary artery disease
2.5.3 The link between rheumatic arthritis and cardiovascular disease
2.6 Treatments for cardiovascular disease
References
CH003.pdf
Chapter 3 Wall quantification and tissue characterization of the coronary artery
3.1 Introduction
3.2 Physics of image acquisition
3.2.1 Image acquisition using optical coherence tomography
3.2.2 Image acquisition using intravascular ultrasound
3.2.3 Comparison of OCT and IVUS
3.3 Tissue characterization
3.3.1 OCT appearance of plaque tissues
3.3.2 Schools of thought on tissue characterization
3.3.3 Characterization using optical properties
3.3.4 Characterization using machine learning
3.3.5 Tissue characterization using deep learning
3.4 A link between carotid and coronary artery disease
3.4.1 Carotid intima–media thickness and CAD
3.4.2 Carotid plaque and CAD
3.4.3 Coronary IMT and carotid atheroma for CAD risk detection
3.4.4 Femoral and carotid IMT for CAD risk detection
3.4.5 Coronary calcium and carotid risk factors for risk detection
3.5 Wall quantification
3.5.1 Lumen measurement
3.5.2 Vessel wall measurement
3.5.3 Fibrous cap measurement
3.5.4 Measurement of calcium
3.5.5 Quantification of macrophages
3.5.6 The role of image registration
3.6 Risk assessment systems
3.7 Discussion
3.7.1 Benchmarking
3.7.2 A note on image acquisition hardware
3.7.3 A note on plaque component quantification
3.7.4 Validation of plaque characterization techniques
3.7.5 A note on the future of OCT
3.8 Conclusion
Appendix A
A.1 Clinical trials
References
CH004.pdf
Chapter 4 Rheumatoid arthritis: its link to atherosclerosis imaging and cardiovascular risk assessment using machine-learning-based tissue characterization
4.1 Introduction
4.2 Search strategy
4.3 Brief description of the pathogensis of rheumatoid arthritis
4.4 Atherosclerosis driven by rheumatoid arthritis
4.5 The role of platelets in atherothrombosis in RA
4.6 The role of amyloidosis in RA
4.7 Traditional CV risk factors in rheumatoid arthritis
4.7.1 Body mass index and physical inactivity
4.7.2 Lipids
4.7.3 Hypertension
4.7.4 Smoking
4.7.5 Insulin resistance and diabetes
4.7.6 Ankle–brachial index and arterial stiffness
4.8 RA-specific CV risk factors in rheumatoid arthritis
4.9 Conventional CV risk algorithms
4.10 Cardiovascular imaging in rheumatoid arthritis
4.10.1 Non-invasive imaging techniques
4.10.2 Invasive imaging techniques: IVUS and OCT
4.11 RA-driven atherosclerotic plaque wall tissue characterization: intelligence paradigm
4.11.1 Machine-learning-based tissue characterization
4.11.2 Deep-learning-based tissue characterization
4.12 Research agenda
4.13 Summary and conclusion
Appendix A
References
CH005.pdf
Chapter 5 A deep-learning fully convolutional network for lumen characterization in diabetic patients using carotid ultrasound: a tool for stroke risk
5.1 Introduction
5.2 Data demographics
5.3 Methodology
5.3.1 Pre-processing
5.3.2 The encoder
5.3.3 The decoder
5.4 Results
5.4.1 Experimental protocol
5.4.2 Experimental results
5.4.3 Performance evaluation
5.5 Discussion
5.5.1 Benchmarking
5.5.2 A short note on skip operation in FCN
5.5.3 A short note on manual tracings of LI borders
5.5.4 Strengths and weaknesses
5.6 Conclusion
Acknowledgment
Appendix A Statistical test results
Appendix B Polyline distance metric
Appendix C Figure-of-merit and precision-of-merit
Appendix D LI-far and LI-near position errors
Appendix E Symbol table
References
CH006.pdf
Chapter 6 Deep-learning strategy for accurate carotid intima–media thickness measurement: an ultrasound study on a Japanese diabetic cohort
6.1 Introduction
6.2 Data demographics and US acquisition
6.3 Methodology
6.3.1 Multiresolution as phase I
6.3.2 DL as phase II
6.3.3 Boundary extraction as phase III
6.3.4 Performance evaluation as phase IV
6.4 Experimental protocol and results
6.4.1 Experimental protocol
6.4.2 Results
6.5 Performance of the DL systems and variability analysis
6.5.1 Comparison of DL against expert manual tracing
6.5.2 Comparison of the DL against the sonographer’s readings
6.5.3 Absolute and signed cIMT error analysis for DL1 and DL2 systems
6.5.4 DL versus previous methods
6.5.5 Interoperator variability of the DL systems: DL1 and DL2
6.5.6 Interobserver variability between the GT systems: GT1 and GT2
6.6 Statistical tests and risk analysis
6.6.1 Four statistical tests
6.6.2 Risk analysis by age
6.6.3 Risk stratification and ROC curves
6.7 Discussion
6.7.1 Benchmarking table
6.7.2 A short note on calibration
6.7.3 A special note on DL optimization
6.7.4 A special note on skip operation
6.7.5 Strengths, weaknesses and extensions
6.7.6 Hardware configuration
6.8 Conclusion
Acknowledgment
Appendix A Polyline distance method
Appendix B Encoder and decoder network
Appendix C LI/MA position errors, cIMT errors and precision-of-merit
Appendix D “
References
CH007.pdf
Chapter 7 Echolucency-based phenotype in carotid atherosclerosis disease for risk stratification of diabetes patients
7.1 Introduction
7.2 Patient demographics and methodology
7.2.1 Patients demographics
7.2.2 Methodology
7.3 Results and statistical analysis
7.3.1 CC analysis of AAGSM and GSMconv against HbA1c
7.3.2 CC between left and right CCA for AAGSM and GSMconv
7.3.3 CC analysis of AAGSM–HbA1c and GSMconv–HbA1c in males and females
7.3.4 Risk stratification based on AAGSM, and HbA1c and ROC analysis
7.3.5 Statistical tests
7.4 Discussion
7.4.1 A note on the HbA1c and AAGSM thresholds for risk stratification
7.4.2 Justification of the δth percentile value during GSMδ measurement
7.4.3 A special note on age-adjustment pre-multiplier (M) selection
7.4.4 A note on the therapeutic implications of AAGSM
7.4.5 Benchmarking against the previous literature
7.4.6 Strengths, weaknesses and applications of AAGSM
7.5 Conclusion
References
CH008.pdf
Chapter 8 Morphologic TPA (mTPA) and composite risk score for moderate carotid atherosclerotic plaque is strongly associated with HbA1c in a diabetes cohort
8.1 Introduction
8.2 Materials and methods
8.2.1 Patient demographics
8.2.2 Six phenotype measurements derived from carotid ultrasound scans
8.2.3 Statistical analysis
8.3 Results
8.3.1 Demographics and clinical characteristics of the patients
8.3.2 Visual display of six phenotypes using AtheroEdge™
8.3.3 Correlation between operators and correlation between cIMT and mTPA for the left, right, and mean of the left and right carotid arteries
8.3.4 Logistic regression for the effect of the six phenotypes on HbA1c for the operator of AtheroEdge™
8.4 Inter-operator variability and statistical tests
8.4.1 Inter-operator variability
8.4.2 Statistical tests
8.5 Discussion
8.5.1 A special note on mTPA and CRS
8.5.2 Benchmarking
8.5.3 A special note on the reproducibility of phenotypes
8.6 Conclusions
Conflict of interest
Contributions
Financial disclosures
Appendix A Box-plots
Appendix B Correlation tables
Appendix C Statistical tests results
Appendix D Abbreviations
References
CH009.pdf
Chapter 9 Plaque tissue morphology-based stroke risk stratification using carotid ultrasound: a polling-based PCA learning paradigm
9.1 Introduction
9.2 Demographics, data collection and preparation
9.2.1 Patient demographics
9.2.2 Data acquisition
9.2.3 Manual wall region extraction for the manual risk assessment system (mRAS)
9.2.4 Modeling the manual LD into two stratification classes: high risk and low risk
9.3 Risk assessment methodology
9.3.1 IMT far and near wall strip extraction
9.3.2 Assessment of stroke risk using a machine-learning system
9.3.3 Texture features
9.3.4 Support vector machine (SVM) and classification
9.3.5 Feature reduction technique using polling-based principal component analysis
9.3.6 Kernel optimization based on the machine-learning paradigm
9.4 Experimental protocol and results
9.4.1 Experiment 1: dominant feature selection and classification accuracy with changing PCA cutoff
9.4.2 Experiment 2: the role of data size in the performance of machine-learning
9.5 Performance evaluation
9.5.1 Precision-of-merit (PoM) analysis
9.5.2 Reliability analysis of the sRAS
9.5.3 Feature retaining power of the sRAS
9.5.4 Stability analysis of the sRAS
9.6 Discussion
9.6.1 About the risk assessment system
9.6.2 Justification for the three kinds of cross-validation protocols
9.6.3 Choice of biomarker (LD versus cIMT)
9.6.4 A note on wall segmentation validation
9.6.5 Benchmarking against the current literature
9.6.6 Summary of our contribution
9.6.7 Strengths, weaknesses and extensions
Appendix A Experimental results
Appendix B Grayscale features
References
CH010.pdf
Chapter 10 Multiresolution-based coronary calcium volume measurement techniques from intravascular ultrasound videos
10.1 Introduction
10.2 Patient demographics and data acquisition
10.2.1 Patient demographics
10.2.2 IVUS data acquisition
10.2.3 Coronary artery data size preparation
10.2.4 Region-of-interest estimation
10.3 Methodology
10.3.1 Overall system
10.3.2 Five multiresolution techniques
10.3.3 Four segmentation methods
10.4 Results
10.4.1 Calcium detection
10.4.2 Volume measurement
10.4.3 Percentage mean time improvement
10.5 Performance evaluation
10.5.1 Multiresolution error metrics against non-multiresolution technique
10.5.2 The mean Jaccard index (JI) and Dice similarity coefficient (DSC)
10.5.3 Manual scoring of detected calcium by a radiologist
10.5.4 Degradation ratio and quality assessment ratio
10.6 Discussion
10.6.1 Our system
10.6.2 Comparison of our down-sampling methods against other methods
10.6.3 A note on gating and registration
10.6.4 Bias correction
10.6.5 A note on time complexity and precision-of-merit
10.6.6 Benchmarking
10.6.7 Strengths, weaknesses and extensions
10.7 Conclusion
Acknowledgments
Funding
Conflicts of interest
Appendix A Tables
Appendix B Mean times of 20 combinations
References
CH011.pdf
Chapter 11 A cloud-based smart lumen diameter measurement tool for stroke risk assessment during multicenter clinical trials
11.1 Introduction
11.2 Materials and methods
11.2.1 Manual lumen diameter reading
11.2.2 Workflow architecture of the AtheroCloud ultrasound system
11.2.3 Engineering design of the AtheroCloud ultrasound system
11.2.4 Two application modes of AtheroCloud: routine mode and pharma trial mode
11.3 Results
11.3.1 Measurements and visualization
11.3.2 Performance evaluation of the AtheroCloud ultrasound system
11.3.3 PoM, FoM, CC and Bland–Altman plots
11.3.4 Cumulative frequency distribution of LD error and TLA error
11.3.5 Receiver operating characteristic
11.3.6 Statistical tests
11.4 Discussion
11.4.1 Our system
11.4.2 Benchmarking AtheroCloud against AtheroEdge™ 2.0
11.4.3 Strengths, weaknesses and extensions
11.5 Conclusion
Acknowledgments
Funding
Conflicts of interest
Appendix A Precision-of-merit and figure-of-merit for AtheroCloud LD measurements
Appendix B Figures
References
CH012.pdf
Chapter 12 A MEMS-based manufacturing technique of vascular bed
12.1 Introduction
12.2 Microstructural anatomy of blood vessels
12.2.1 Arteries and veins
12.2.2 Capillaries
12.3 Modeling of blood vessels as a microsystem
12.3.1 Acoustic wave mechanosensors
12.3.2 Pressure mechanosensors
12.3.3 Microvalves and micropumps
12.4 Scaling laws of miniaturized blood vessels
12.4.1 Scaling in geometry
12.4.2 Scaling in fluid dynamics
12.5 Microfabrication of blood vessels
12.5.1 Soft lithography techniques
12.5.2 Self-assembly techniques
12.5.3 Sputtering techniques
12.6 Microvessel design
12.6.1 Design consideration
12.6.2 Mechanical design of a balloon angioplasty pressure sensor using finite element methods
12.7 Conclusion
References