Multi-modality Cardiac Imaging

Patrick Clarysse is Researcher for the French National Center for Scientific Research (CNRS) working at the CREATIS laboratory in Lyon, France. His research works focus on multi-dimensional medical image processing and modeling with primary applications in cardiovascular imaging.Denis Friboulet is Professor at INSA-Lyon, France and works at the CREATIS laboratory in Lyon, France. His research interests include advanced signal and image processing with application to 2D/3D echocardiographic imaging.

• PREFACE xiiiACKNOWLEDGMENTS xvINTRODUCTION xviiPART 1. METHODOLOGICAL BASES 1CHAPTER 1. EXTRACTION AND SEGMENTATION OF STRUCTURES IN IMAGE SEQUENCES 3Olivier BERNARD, Patrick CLARYSSE, Thomas DIETENBECK, Denis FRIBOULET, Stéphanie JEHAN-BESSON and Jérome POUSIN1.1. Problematics 31.2. Overview of segmentation methods 31.3. Summary of the different classes of deformable models 61.3.1. Non-energy approaches 71.3.2. Energy-based approaches 81.4. Deformable templates 111.4.1. Elastic deformable template principle 121.4.2. Dynamic elastic deformable template 141.4.3. Elastic deformable template and modal analysis 151.4.4. The elastic deformable template in practice 151.5. Variational active contours 171.5.1. Active contour representations 171.5.2. Energy functional 211.5.3. Obtaining the evolution equation 261.5.4. Level set digital implementation 341.6. Integration of a priori constraints in the formalism of variational contours 351.6.1. Shape a priori 361.6.2. Motion a priori 381.7. Implementation examples in cardiac imaging 441.7.1. Echographic imaging: choice of the data fitting term 441.7.2. Example of 3D echocardiography image segmentation 461.7.3. Example of 2D echocardiography image segmentation 481.8. Conclusion 501.9. Bibliography 52CHAPTER 2. MOTION ESTIMATION AND ANALYSIS 65Patrick CLARYSSE and Jérome POUSIN2.1. Problematics 652.2. Problem formulation 662.3. Transport methods 672.3.1. Optical flow 682.3.2. Motion estimation seen as an optimal transport problem 702.4. Probabilistic approaches 742.5. Image registration 762.5.1. Transformation 772.5.2. Similarity function 782.5.3. Optimization 782.5.4. Practical considerations 792.6. Local methods 792.6.1. Block or primitive-matching 792.6.2. Least-square estimation 812.7. Hybrid methods 812.7.1. Power spectrum-based methods 822.7.2. Spatiotemporal description 822.8. Phase-based methods 842.8.1. Fleet and Jepson’s method 852.8.2. Analytic and monogenic signal 862.8.3. Harmonic phase methods 882.9. Registration and motion estimation in a sequence of images 892.9.1. Lagrangian description 892.9.2. Eulerian description 912.9.3. Strategies for the estimation in sequence 912.10. Evaluation of motion estimation methods 922.11. Conclusion 952.12. Bibliography 95CHAPTER 3. POST-PROCESSING AND ANALYSIS OF DYNAMIC MAGNETIC RESONANCE IMAGES FOR MYOCARDIAL PERFUSION QUANTIFICATION 103Bruno NEYRAN and Magalie VIALLON3.1. Introduction 1033.2. Dynamic measurement of perfusion with contrast agents: reminder about the MRI sequences and the different contrast agents used 1073.2.1. Brief reminder about cardiac perfusion MRI sequences 1073.2.2. MRI signal conversion/tracer concentration 1073.2.3. Different clinical-candidate contrast agents 1083.3. Motion correction and contour segmentation of the myocardium: important preprocessing prior to quantitative analysis 1093.3.1. Dynamic image registration 1093.3.2. Automatic contour extraction 1093.4. Semi-quantitative perfusion analysis: calculation of relative parameters depending on the injection of the contrast medium 1103.4.1. Semi-quantitative perfusion parameters 1103.4.2. Heuristic modeling using a varied gamma function 1123.4.3. Heuristic modeling with a bi-exponential function 1143.4.4. Heuristic modeling with the Moate model 1153.5. Absolute parameters independent of the contrast agent injection (taking account of the arterial input): pharmacokinetic modeling 1173.5.1. General studies: tracer kinetics theory 1183.5.2. Identification of the residual function 1273.5.3. Identification of the discrete residual function 1293.6. Conclusion 1333.7. Bibliography 135CHAPTER 4. TENSOR DECOMPOSITION OF A DYNAMIC SEQUENCE OF IMAGES INTO SIMPLE ELEMENTS  141Frédérique FROUIN and Claire PELLOT-BARAKAT4.1. Problematics 1414.2. Panorama of methods for the quantitative analysis of dynamic image sequences 1434.2.1. Regions of interest method 1434.2.2. Parametric imaging methods 1444.2.3. Movement analysis methods 1454.2.4. Tensor decomposition of a sequence of images into simple elements 1454.3. Tensor decomposition methods of an image sequence into simple elements 1464.3.1. Notations and decomposition principle 1464.3.2. Orthogonal decomposition of an image sequence 1474.3.3. Decomposition into simple elements 1484.4. Specifications for radiotracer or contrast medium monitoring 1494.4.1. Proposed approach objectives and associated constraints definition 1494.4.2. Components estimation principle 1494.4.3. Example of tensor decomposition into simple elements in myocardial perfusion studies 1524.4.4. Limitations of the proposed approach 1534.4.5. Clinical applications of the tensor decomposition into simple elements for cardiac imaging 1554.5. Specifications for the study of cardiac motion 1564.5.1. Proposed approach objectives and associated constraint definition 1564.5.2. Tensor decomposition method solution 1574.5.3. Tensor decomposition model extensions 1604.5.4. Clinical applications and perspectives 1644.6. Conclusion 1654.7. Bibliography 166PART 2. APPLICATION EXAMPLES 169CHAPTER 5. EVALUATION OF CARDIAC STRUCTURE SEGMENTATION IN CINE MAGNETIC RESONANCE IMAGING 171Alain LALANDE, Mireille GARREAU and Frédérique FROUIN5.1. Context: significance of the automatic segmentation of the cardiac structures 1715.1.1. Cine MRI in short-axis orientation 1715.1.2. Left ventricle and right ventricle 1725.2. Evaluation necessity 1755.2.1. The place of evaluation 1755.2.2. Analytic and empirical methods 1765.3. Empirical evaluation methods 1775.4. Visual evaluation methods 1795.5. Supervised methods 1805.5.1. The definition of a reference 1805.5.2. Creation of an expert database 1835.5.3. Evaluation criterion: edge-based approaches 1845.5.4. Evaluation criteria: region-based approaches 1885.5.5. Supervised methods for the estimation of a clinical parameter 1925.5.6. ROC curves 1935.5.7. Comparison of the supervised methods 1945.5.8. Limitations of the supervised methods 1955.6. Non-supervised evaluation methods 1985.6.1. Unsupervised methods relying on region- or edge-based descriptors 1985.6.2. Methods using a clinical parameter 2025.6.3. Estimation methods of a reference segmentation 2045.6.4. Difficulties in unsupervised methods 2055.7. Conclusion 2055.8. IMPEIC and MEDIEVAL working groups 2075.9. Bibliography 209CHAPTER 6. PHASE-BASED HEART MOTION ESTIMATION IN MULTIMODALITY CARDIAC IMAGING 217Martino ALESSANDRINI, Adrian BASARAB, Olivier BERNARD and Philippe DELACHARTRE6.1. Phase images 2186.1.1. Multidimensional analytic signals 2186.1.2. Monogenic signal 2196.2. Optical flow motion estimation on the phase of the two single-orthant analytic signals and using a deformable mesh: application to cardiac MRI sequences 2216.2.1. Optical flow method applied to spatial phase images 2236.2.2. Parametric modeling of local motion 2266.2.3. Trajectory estimation 2286.2.4. Results 2306.2.5. Conclusion 2356.3. Motion estimation by optical flow from the monogenic phase using a local affine model and multiscale analysis – application to ultrasonic cardiac sequences 2366.3.1. Affine model 2376.3.2. Multiscale choice of the window size 2386.3.3. Iterative refinement of the displacement 2386.4. Bibliography 244CHAPTER 7. CARDIAC MOTION ANALYSIS IN TAGGED MRI 247Patrick CLARYSSE and Pierre CROISILLE7.1. Motion quantification by the SinMod method 2487.2. Processing pipeline and features of the software inTag 2507.2.1. Data and input parameters 2517.2.2. Motion field estimation 2517.2.3. LV contour extraction 2527.2.4. LV motion and deformation analysis 2527.3. Perspectives 2547.4. Bibliography 254CHAPTER 8. LEFT VENTRICLE MOTION ESTIMATION IN COMPUTED TOMOGRAPHY IMAGING 257Antoine SIMON, Mireille GARREAU, Régis DELAUNAY, Dominique BOULMIER, Erwan DONAL and Christophe LECLERCQ8.1. Introduction 2578.1.1. Clinical problem and objectives 2578.1.2. Technological choice: cardiac CT imaging 2588.1.3. State of the art and method positioning 2598.2. Surface matching method 2628.2.1. Surface segmentation and reconstruction stage 2628.2.2. Surface–surface matching 2638.3. Surface–surface approach evaluation 2678.3.1. Simulated data 2678.3.2. Real data 2708.4. Surface–surface approach conclusion 2788.5. Surface and volume matching method: surface–volume approach 2788.6. Surface–volume approach evaluation 2808.6.1. Simulated data 2808.6.2. Real data 2838.7. Conclusion 2858.8. Acknowledgments 2878.9. Bibliography 287PART 3 . TOWARD PATIENT-SPECIFIC CARDIOLOGY 293CHAPTER 9. PERSONALIZATION OF ELECTROMECHANICAL MODELS OF THE CARDIAC VENTRICULAR FUNCTION BY HETEROGENEOUS CLINICAL DATA ASSIMILATION 295Stephanie MARCHESSEAU, Maxime SERMESANT, Florence BILLET, Hervé DELINGETTE and Nicholas AYACHE9.1. Introduction 2959.2. Anatomy and electrophysiology personalization from clinical data 2989.2.1. Personalization of the heart and the tissue structure anatomy 2989.2.2. Cardiac electrophysiology personalization 3009.3. Heart mechanics modeling 3029.3.1. Modeling of the Bestel–Clément–Sorine electromechanical coupling 3029.3.2. Blood flow modeling 3049.3.3. Other boundary conditions 3059.3.4. Discussion about this model 3069.4. Image data processing: cardiac kinematics personalization 3069.4.1. Metrics for the comparison between observed and simulated motion 3079.4.2. Data time interpolation 3079.4.3. Deformable models approach 3089.4.4. Data displacement case 3109.4.5. Velocity data case 3119.4.6. Results with cine-MRI data 3119.4.7. Results from dynamic CT data 3129.5. Calibration of the mechanical parameters from global data 3139.5.1. Available data description 3149.5.2. Unscented transform calibration 3159.5.3. Calibration results with healthy volunteers 3179.5.4. Calibration results with pathological cases 3179.6. Mechanical personalization by variational data assimilation 3189.6.1. Variational approach on a simplified model 3209.6.2. Application to synthetic cases 3219.6.3. Application to clinical cases 3229.6.4. Sequential approach on full model 3229.7. Conclusion 3239.8. Bibliography 324CONCLUSION 331APPENDIX 1 335APPENDIX 2 339LIST OF AUTHORS 343INDEX 347