Bayesian and GrAphical Models for Biomedical Imaging : First International Workshop, BAMBI 2014, Cambridge, MA, USA, September 18, 2014, Revised Selected Papers

This book constitutes the refereed proceedings of the First International Workshop on Bayesian and grAphical Models for Biomedical Imaging, BAMBI 2014, held in Cambridge, MA, USA, in September 2014 as a satellite event of the 17th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2014. The 11 revised full papers presented were carefully reviewed and selected from numerous submissions with a key aspect on probabilistic modeling applied to medical image analysis. The objectives of this workshop compared to other workshops, e.g. machine learning in medical imaging, have a stronger mathematical focus on the foundations of probabilistic modeling and inference. The papers highlight the potential of using Bayesian or random field graphical models for advancing scientific research in biomedical image analysis or for the advancement of modeling and analysis of medical imaging data. Bayesian and grAphical Models for Biomedical Imaging Preface 5 Organization 6 Table of Contents 8 N3 Bias Field Correction Explained as a Bayesian Modeling Method 10 1 Introduction 10 2 Methods 11 2.1 The N3 Method in Its Practical Implementation 11 2.2 EM-Based Bias Field Estimation 13 2.3 N3 as an Approximate MAP Parameter Estimator 15 3 Experiments 16 4 Results 18 5 Discussion 20 References 21 A Bayesian Approach to Distinguishing Interdigitated Muscles in the Tongue from Limited Diffusion Weighted Imaging 22 1 Introduction 22 2 Methods 23 2.1 Multi-tensor Model with a Fixed Tensor Basis 23 2.2 Mixture Fraction Estimation with Prior Knowledge 24 3 Experiments 25 3.1 Digital Phantom 25 3.2 In Vivo Tongue Diffusion Data 28 4 Discussion 30 5 Conclusion 32 References 33 Segmentation and Tracking of E. coli 34 1 Introduction 34 2 Microscopic Setup and Data Preprocessing 36 3 Segmentation Methods 37 3.1 Thresholding and Component Trees (CT) 37 3.2 Parametric Max-Flow (PMF) 38 3.3 Parametric Max-Flow and Random Forest (PMFRF) 38 4 A Graphical Model for Segmentation and Tracking 39 4.1 Costs 40 4.2 Constraints 41 4.3 Eliminating Segmentation Variables 42 4.4 Finding The Globally Optimal Solution 43 5 Results 43 6 Summary and Discussion 44 References 45 Physiologically Informed Bayesian Analysis of ASL fMRI Data 46 1 Introduction 46 2 A Physiologically Informed ASL/BOLD Link 47 2.1 The Extended Balloon Model 47 2.2 Physiological Linear Relationship between Response Functions 49 3 Bayesian Hierarchical Model for ASL Data Analysis 50 4 A Physiologically Informed 2-steps Inference Procedure 52 4.1 Hemodynamics Estimation Step M1 52 4.2 Perfusion Response Estimation Step M2 53 5 Simulation Results 53 6 Real Data Results 55 7 Discussion and Conclusion 55 References 56 Bone Reposition Planning for Corrective Surgery Using Statistical Shape Model: Assessment of Differential Geometrical Features 58 1 Introduction 58 2 SSM Based Planning 60 2.1 Fitting of the SSM to Two Bone Segements 60 2.2 Probability Distribution Functions for Shape Validity and Scaling 61 2.3 Probability Distribution Model to Compare Shapes 61 3 Experiments 62 3.1 Data 62 3.2 Evaluating the SSM 63 3.3 Evaluating Modes of Variation 64 3.4 Including the Geometrical Features 64 3.5 Accuracy of Bone Repositioning 64 4 Discussion 66 5 Appendix 67 An Inference Language for Imaging 70 1 Introduction 70 2 Methods 71 2.1 The Modeling Language 72 2.2 The Inference Engine 75 3 Motion-aware Positron Emission Tomography 78 4 Conclusion 80 5 Download 80 References 81 An MRF-Based Discrete Optimization Framework for Combined DCE-MRI Motion Correction and Pharmacokinetic Parameter Estimation 82 1 Introduction 82 2 Methods 84 2.1 Data Cost Calculation Using Pharmacokinetic Model Prediction 84 2.2 Optimization on the Reduced 4D Graph 85 3 Results 86 3.1 Algorithm Evaluation on Synthetic Data 86 3.2 Pharmacokinetic Modelling and Motion Correction on DCE-MRI Images of Rectal Cancer 89 4 Discussion and Conclusion 91 5 Future Work 91 References 92 Learning Imaging Biomarker Trajectories from Noisy Alzheimer's Disease Data Using a Bayesian Multilevel Model 94 1 Introduction 94 2 Data and Methods 95 3 Results 98 4 Discussion 101 References 102 Four Neuroimaging Questions that P-Values Cannot Answer (and Bayesian Analysis Can) 104 1 Introduction 104 2 Brain Imaging Is Free of Type I and Type II Errors 104 3 Type S Errors: How Confident Are We That Our Finding Is not Opposite to the Truth? 105 4 Type M Errors: Can the True Effect Be Much Smaller than What We Observed? 108 5 Do Patients and Controls Have Similar Brains? 109 6 What Is the Probability That the Patient Has the Disease? 111 7 Discussion 112 8 Conclusion 114 References 115 Spherical Topic Models for Imaging Phenotype Discovery in Genetic Studies 116 1 Introduction 116 2 Topic Modeling for Feature Extraction 118 3 From Image Features to Genetic Markers 120 4 Experiments 121 5 Conclusion 125 References 125 A Generative Model for Automatic Detection of Resolving Multiple Sclerosis Lesions 127 1 Introduction 127 2 Method 129 2.1 Bayesian Formulation 129 3 Experiments 131 3.1 Data Sets 131 3.2 Validation 135 4 Discussion 137 References 138 Author Index 139 Front Matter....Pages - N3 Bias Field Correction Explained as a Bayesian Modeling Method....Pages 1-12 A Bayesian Approach to Distinguishing Interdigitated Muscles in the Tongue from Limited Diffusion Weighted Imaging....Pages 13-24 Optimal Joint Segmentation and Tracking of Escherichia Coli in the Mother Machine....Pages 25-36 Physiologically Informed Bayesian Analysis of ASL fMRI Data....Pages 37-48 Bone Reposition Planning for Corrective Surgery Using Statistical Shape Model: Assessment of Differential Geometrical Features....Pages 49-60 An Inference Language for Imaging....Pages 61-72 An MRF-Based Discrete Optimization Framework for Combined DCE-MRI Motion Correction and Pharmacokinetic Parameter Estimation....Pages 73-84 Learning Imaging Biomarker Trajectories from Noisy Alzheimer’s Disease Data Using a Bayesian Multilevel Model....Pages 85-94 Four Neuroimaging Questions that P-Values Cannot Answer (and Bayesian Analysis Can)....Pages 95-106 Spherical Topic Models for Imaging Phenotype Discovery in Genetic Studies....Pages 107-117 A Generative Model for Automatic Detection of Resolving Multiple Sclerosis Lesions....Pages 118-129 Back Matter....Pages -