Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries : 6th International Workshop, BrainLes 2020, Held in Conjunction with MICCAI 2020, Lima, Peru, October 4, 2020, Revised Selected Papers, Part I

This two-volume set LNCS 12658 and 12659 constitutes the thoroughly refereed proceedings of the 6th International MICCAI Brainlesion Workshop, BrainLes 2020, the International Multimodal Brain Tumor Segmentation (BraTS) challenge, and the Computational Precision Medicine: Radiology-Pathology Challenge on Brain Tumor Classification (CPM-RadPath) challenge. These were held jointly at the 23rd Medical Image Computing for Computer Assisted Intervention Conference, MICCAI 2020, in Lima, Peru, in October 2020.* The revised selected papers presented in these volumes were organized in the following topical sections: brain lesion image analysis (16 selected papers from 21 submissions); brain tumor image segmentation (69 selected papers from 75 submissions); and computational precision medicine: radiology-pathology challenge on brain tumor classification (6 selected papers from 6 submissions). *The workshop and challenges were held virtually. Preface Organization Contents – Part I Contents – Part II Invited Papers Glioma Diagnosis and Classification: Illuminating the Gold Standard 1 Features of Infiltrating Gliomas 1.1 Introduction 1.2 The Importance of IDH Mutations in Infiltrating Gliomas 1.3 Diagnostic Format 1.4 Refining Diagnoses and Improving Patient Care References Multiple Sclerosis Lesion Segmentation - A Survey of Supervised CNN-Based Methods 1 Introduction 2 Review of Methods 2.1 Data Pre-processing 2.2 Data Representation 2.3 Data Preparation 2.4 Network Architecture 2.5 Multiple Modalities, Timepoints, Views and Scales 2.6 Loss Functions and Regularization 2.7 Implementation 2.8 Prediction and Post-processing 2.9 Transfer Learning and Domain Adaptation 2.10 Methods for Subtypes of MS Lesions 3 Comparison of Experiments and Results 3.1 Datasets 3.2 Evaluation Metrics 3.3 Results 4 Conclusion References Computational Diagnostics of GBM Tumors in the Era of Radiomics and Radiogenomics 1 Introduction 2 Patient Prognosis 3 Intratumor Heterogeneity and Tumor Recurrence 4 Radiogenomics 5 Current Challenges and Future Directions 6 Conclusion References Brain Lesion Image Analysis Automatic Segmentation of Non-tumor Tissues in Glioma MR Brain Images Using Deformable Registration with Partial Convolutional Networks 1 Introduction 2 Method 2.1 Tumor Segmentation Network 2.2 Tumor Region Recovery Network 3 Results 3.1 Data and Experimental Setting 3.2 Evaluation of Image Recovery 3.3 Evaluation of Image Registration 4 Conclusion References Convolutional Neural Network with Asymmetric Encoding and Decoding Structure for Brain Vessel Segmentation on Computed Tomographic Angiography 1 Introduction 2 Data Acquisition and Preprocessing 3 Method 3.1 Asymmetric Encoding and Decoding-Based Convolutional Neural Net 3.2 Centerline Loss Construction 3.3 Network Training 4 Results 5 Conclusions References Volume Preserving Brain Lesion Segmentation 1 Introduction 2 Methods 2.1 Background: Existing Inequality Volume Constraints 2.2 Volume Constraints for Fully Supervised Settings 2.3 Implementation Details 3 Results 3.1 Dataset 3.2 Evaluation 4 Discussion 5 Conclusion References Microstructural Modulations in the Hippocampus Allow to Characterizing Relapsing-Remitting Versus Primary Progressive Multiple Sclerosis 1 Introduction 2 Materials and Methods 2.1 Study Participants and MRI Acquisition 2.2 Signal Modelling and Microstructural Indices 2.3 Image Preprocessing 2.4 Features Extraction 2.5 Statistical Analysis 3 Results 4 Discussion 5 Conclusions References Symmetric-Constrained Irregular Structure Inpainting for Brain MRI Registration with Tumor Pathology 1 Introduction 2 Methods 2.1 Global Perception Inference 2.2 Context-Aware Patch Swapping 2.3 Feature-to-Image Translator 2.4 Quasi-Symmetry Constraint 3 Experiments and Results 4 Conclusion References Multivariate Analysis is Sufficient for Lesion-Behaviour Mapping 1 Introduction 2 Multivariate Methods Considered 3 Causal Analysis of the Problem 4 Experiments 5 Discussion 5.1 Outlook 5.2 Future Work References Label-Efficient Multi-task Segmentation Using Contrastive Learning 1 Introduction 2 Methods 2.1 Encoder-Decoder Network with Regularization Branches 3 Experiments and Results 4 Discussion and Conclusion References Spatio-Temporal Learning from Longitudinal Data for Multiple Sclerosis Lesion Segmentation 1 Introduction 2 Methodology 2.1 Baseline Longitudinal Network 2.2 Multitask Learning with Deformable Registration 3 Experiment Setup 3.1 Datasets and Preprocessing 3.2 Implementation Details 3.3 Evaluation Metrics 3.4 Method Comparisons 4 Results and Discussion 5 Conclusion References MMSSD: Multi-scale and Multi-level Single Shot Detector for Brain Metastases Detection 1 Introduction 2 Methodology 2.1 Multi-scale Feature Maps for BM Detection 2.2 Multi-level Feature Fusion Modules 3 Experiments and Results 3.1 Dataset and Data Processing 3.2 Training 3.3 Experiment Results 3.4 Discussion 4 Conclusion References Unsupervised 3D Brain Anomaly Detection 1 Introduction 2 Methods 3 Results and Discussion 4 Conclusion References Assessing Lesion Segmentation Bias of Neural Networks on Motion Corrupted Brain MRI 1 Introduction 2 Methods 3 Experiments 4 Results and Discussion 5 Conclusion and Future Work References Estimating Glioblastoma Biophysical Growth Parameters Using Deep Learning Regression 1 Introduction 2 Methods 2.1 Data 2.2 Pre-processing 2.3 Network Topology 2.4 Experimental Design 2.5 Evaluation Metric 3 Results 4 Discussion References Bayesian Skip Net: Building on Prior Information for the Prediction and Segmentation of Stroke Lesions 1 Introduction 2 Materials and Methods 2.1 Data 2.2 Data Pre-processing 2.3 Network Architecture 2.4 Experimental Setup 3 Results 4 Discussion and Conclusion References Brain Tumor Segmentation Brain Tumor Segmentation Using Dual-Path Attention U-Net in 3D MRI Images 1 Introduction 2 Method 2.1 Backbone Architecture for Segmentation 2.2 Dual Residual Block 2.3 Dual-Path Attention (DPA) Block for Tumor Segmentation 2.4 Loss Function 3 Experiments 3.1 Datasets Detail 3.2 Preprocessing Methods 3.3 Evaluating Metrics 3.4 Experimental Results 4 Discussion 5 Conclusion References Multimodal Brain Image Analysis and Survival Prediction Using Neuromorphic Attention-Based Neural Networks 1 Introduction 2 Methods 2.1 Neuromorphic Neural Network Inspired by Visual Cortex 2.2 Neuromorphic Pre-processing for Convolutional Neural Networks with Neuromorphic Attention-Based Learner 2.3 Feature Selection for Survival Days Prediction 3 Result and Discussion 3.1 Tumor Segmentation 3.2 Overall Survival Prediction 4 Conclusion References Context Aware 3D UNet for Brain Tumor Segmentation 1 Introduction 2 Proposed Method 2.1 Dense Blocks 2.2 Residual-Inception Blocks 3 Experimental Results 3.1 Dataset 3.2 Implementation Details 3.3 Qualitative Analysis 3.4 Quantitative Analysis 4 Discussion and Conclusion References Brain Tumor Segmentation Network Using Attention-Based Fusion and Spatial Relationship Constraint 1 Introduction 2 Method 2.1 Multi-modal Tumor Segmentation Network 2.2 Spatial-Channel Fusion Block (SCFB) 2.3 Spatial Relationship Constraint 3 Experiment 3.1 Dataset 3.2 Implementations 3.3 Results 4 Conclusion References Modality-Pairing Learning for Brain Tumor Segmentation 1 Introduction 2 Methods 3 Experiments 3.1 Dataset 3.2 Experimental Settings 3.3 Evaluation Metrics 3.4 Experimental Results 4 Discussion and Conclusion References Transfer Learning for Brain Tumor Segmentation 1 Introduction 2 Method 2.1 Extending the AlbuNet Architecture 2.2 Loss Function 2.3 Choice of Hyperparameters 2.4 Preprocessing and Data Augmentation 2.5 Prediction 3 Evaluation 3.1 Extended AlbuNet with and Without Pretraining 3.2 Clinical Dataset of the Syrian-Lebanese Hospital 3.3 Final Results 4 Conclusion References Efficient Embedding Network for 3D Brain Tumor Segmentation 1 Introduction 2 Method 2.1 Data Pre-processing 2.2 Encoder Branch 2.3 Decoder Branch 2.4 Loss 2.5 Training 3 Results 4 Conclusion References Segmentation of the Multimodal Brain Tumor Images Used Res-U-Net 1 Introduction 2 Method 2.1 Pre-processing 2.2 Architecture 3 Experiments 3.1 Dataset 3.2 Setup 3.3 Evaluation 3.4 Result 4 Conclusion References Vox2Vox: 3D-GAN for Brain Tumour Segmentation 1 Introduction 1.1 Related Works 2 Method 2.1 Data 2.2 Image Pre-processing and Augmentation 2.3 Model Architecture 2.4 Losses 2.5 Optimization and Regularization 2.6 Model Ensembling and Post-processing 3 Results 4 Conclusions References Automatic Brain Tumor Segmentation with Scale Attention Network 1 Introduction 2 Related Work 3 Methods 3.1 Overall Network Structure 3.2 Encoding Pathway 3.3 Decoding Pathway 3.4 Scale Attention Block 3.5 Implementation 4 Results 5 Summary References Impact of Spherical Coordinates Transformation Pre-processing in Deep Convolution Neural Networks for Brain Tumor Segmentation and Survival Prediction 1 Introduction 2 Brain Tumor Segmentation 2.1 Background 2.2 Spherical Coordinates Transformation 2.3 Post-processing 2.4 Segmentation Results 3 Prediction of Patient OS 3.1 LesionEncoder Framework 3.2 Tweedie Regressor 3.3 Performance Evaluation 3.4 OS Prediction Results 4 Discussion 5 Conclusion References Overall Survival Prediction for Glioblastoma on Pre-treatment MRI Using Robust Radiomics and Priors 1 Introduction 2 Materials and Methods 2.1 Data 2.2 Segmentation and Normalization 2.3 Radiomic Features 2.4 Dimensionality Reduction 2.5 Priors 2.6 Survival Time Transform 3 Results 3.1 Age, Shape and Position Model 3.2 Influence of Priors 3.3 Influence of Target Variable Transform 3.4 Final Model Selection and Challenge Outcome 4 Conclusion References Glioma Segmentation Using Encoder-Decoder Network and Survival Prediction Based on Cox Analysis 1 Introduction 2 Segmentation Method 2.1 Encoder 2.2 Decoder 3 Survival Prediction Method 3.1 Feature Extraction 3.2 Feature Selection 3.3 Regression Framework 4 Experiments 4.1 Data 4.2 Results 5 Conclusion References Brain Tumor Segmentation with Self-ensembled, Deeply-Supervised 3D U-Net Neural Networks: A BraTS 2020 Challenge Solution 1 Introduction 1.1 Clinical Overview 1.2 Multimodal Brain Tumor Segmentation Challenge 2020 2 Methods 2.1 Neural Network Architecture 2.2 Loss Function 2.3 Image Pre-processing 2.4 Data Augmentation Techniques 2.5 Training Details 2.6 Inference 2.7 Ablation Study for Pipeline A 3 Results 3.1 Online Validation Dataset 3.2 Testing Dataset 4 Discussion 5 Conclusion References Brain Tumour Segmentation Using a Triplanar Ensemble of U-Nets on MR Images 1 Introduction 2 Materials and Methods 2.1 Data 2.2 Preprocessing 2.3 CNN Architecture 2.4 Post-processing 2.5 Implementation Details 2.6 Data Augmentation 2.7 Performance Evaluation Metrics 3 Experiments and Results 4 Discussion and Conclusions References MRI Brain Tumor Segmentation Using a 2D-3D U-Net Ensemble 1 Task 1: Brain Tumor Segmentation in MRI Scans 1.1 Materials and Methods 1.2 2D Networks 1.3 3D Network 1.4 2D-3D Ensembling 1.5 Post-processing 1.6 Results 2 Task 2: Survival Task 2.1 Materials and Methods 2.2 Data Sources 2.3 Regions of Interest (ROIs) 2.4 Feature Extraction 2.5 Feature Selection 2.6 Training Data and Models 2.7 Results 2.8 Conclusions References Multimodal Brain Tumor Segmentation and Survival Prediction Using a 3D Self-ensemble ResUNet 1 Introduction 2 Method 2.1 Brain Tumor Segmentation 2.2 Survival Prediction 3 Materials and Pre-processing 3.1 Data 3.2 Pre-processing 4 Experiments and Results 4.1 Hyper-parameter Setting 4.2 Training Stage 4.3 Online Evaluation 5 Conclusion References MRI Brain Tumor Segmentation and Uncertainty Estimation Using 3D-UNet Architectures 1 Introduction 2 Related Work 2.1 Semantic Segmentation 2.2 Uncertainty 3 Method 3.1 Dataset Statistics 3.2 Data Pre-processing and Augmentation 3.3 Sampling Strategy 3.4 Loss 3.5 Network Architecture 3.6 Post-processing 3.7 Uncertainty 4 Results 4.1 Segmentation 4.2 Uncertainty 5 Discussion and Conclusions References Utility of Brain Parcellation in Enhancing Brain Tumor Segmentation and Survival Prediction 1 Introduction 2 Materials and Methods 2.1 Dataset 2.2 Tumor Segmentation 2.3 Survival Prediction 2.4 Implementation Details 3 Results 3.1 Segmentation Results 3.2 Survival Prediction Results 4 Discussion and Conclusion References Uncertainty-Driven Refinement of Tumor-Core Segmentation Using 3D-to-2D Networks with Label Uncertainty 1 Introduction 2 Heteroscedastic Label-Flip Loss and Focal KL-divergence 2.1 Ensembling and Label Uncertainty 3 Application to Brain Tumor Segmentation 3.1 Data Preparation and Homogenization 3.2 The DeepSCAN Architecture with Attention 4 Uncertainty-Based Filtering 4.1 Results 5 Uncertainty Challenge 6 Survival Prediction 7 Results and Conclusions References Multi-decoder Networks with Multi-denoising Inputs for Tumor Segmentation 1 Introduction 2 Methods 2.1 Encoder Network 2.2 Multi-decoder Networks 2.3 Squeeze-and-Excitation Block 2.4 Denoising the Inputs 2.5 Preprocessing and Augmentation 2.6 Post-processing 2.7 Task 3: Quantification of Uncertainty in Segmentation 3 Experiments 3.1 Implementation Details and Training 3.2 Loss 3.3 Optimization 4 Results and Discussion 5 Conclusion References MultiATTUNet: Brain Tumor Segmentation and Survival Multitasking 1 Introduction 2 Data 3 Methodology 3.1 T-EDet 3.2 UNet 3D 3.3 Multitask with MultiATTUNet 4 Results 4.1 Test: Official Validation 4.2 Qualitative Results 4.3 Test 2: Official Test 5 Discussion and Future Work 6 Conclusion References A Two-Stage Cascade Model with Variational Autoencoders and Attention Gates for MRI Brain Tumor Segmentation 1 Introduction 2 Method 2.1 The First-Stage Network: Asymmetrical U-Net with a VAE Branch 2.2 The Second-Stage Network: Attention-Gated Asymmetrical U-Net with the VAE Branch 2.3 Loss Function 3 Expriment 3.1 Data Description 3.2 Implementation Details 4 Results 4.1 Quantitative Results 4.2 Attention Map 5 Concluding Remarks References Multidimensional and Multiresolution Ensemble Networks for Brain Tumor Segmentation 1 Introduction 2 Materials and Methods 2.1 Data and Preprocessing 2.2 Network Architecture 3 Results 3.1 Segmentation 3.2 Survival Prediction 4 Discussion 5 Conclusion References Cascaded Coarse-to-Fine Neural Network for Brain Tumor Segmentation 1 Introduction 2 Methods 2.1 Cascaded Framework 2.2 Coarse-to-Fine Networks 2.3 Logarithmic Dice 3 Experiments and Preliminary Results 4 Conclusion References Low-Rank Convolutional Networks for Brain Tumor Segmentation 1 Introduction 2 Related Work 3 Method 3.1 Data Preprocessing and Augmentation 3.2 Network Architecture 3.3 Low-Rank Convolution 3.4 Loss Function 3.5 Optimization 3.6 Postprocessing 4 Experiments and Results 5 Conclusion References Automated Brain Tumour Segmentation Using Cascaded 3D Densely-Connected U-Net 1 Introduction 2 Methodology 2.1 Dataset Preprocessing 2.2 Network Architecture 2.3 Training Procedure 2.4 Tumour Localization 2.5 Post-processing 3 Experiments and Results 4 Conclusion References Segmentation then Prediction: A Multi-task Solution to Brain Tumor Segmentation and Survival Prediction 1 Introduction 2 Method 2.1 Segmentation Module 2.2 Survival Prediction Module 2.3 Hybrid Loss Function 2.4 Testing 3 Dataset 4 Experiments and Results 4.1 Data Processing 4.2 Experiment Settings 4.3 Results of Segmentation 4.4 Results of Survival Prediction 5 Conclusion References Enhancing MRI Brain Tumor Segmentation with an Additional Classification Network 1 Introduction 2 Methods 2.1 Bi-directional Feature Pyramid Network 2.2 Nested U-Net 2.3 Classification Branch 2.4 Losses 3 Experiments 3.1 Data Pre-processing and Augmentation 3.2 Training Details 3.3 Inference 4 Results 5 Conclusion References Self-training for Brain Tumour Segmentation with Uncertainty Estimation and Biophysics-Guided Survival Prediction 1 Introduction 2 Methods 2.1 Tumour Sub-region Segmentation 2.2 Uncertainty Measure in Segmentation 2.3 Invasiveness Feature Extraction 2.4 Prognostic Model 3 Results 3.1 Segmentation Performance 3.2 Uncertainty Prediction 3.3 Survival Prediction 4 Discussion 5 Conclusion References Author Index