Machine Learning in Clinical Neuroimaging: 4th International Workshop, MLCN 2021, Held in Conjunction with MICCAI 2021, Strasbourg, France, September ... (Lecture Notes in Computer Science, 13001)

Preface
Organization
Contents
Computational Anatomy
Unfolding the Medial Temporal Lobe Cortex to Characterize Neurodegeneration Due to Alzheimer’s Disease Pathology Using Ex vivo Imaging
	1 Introduction
	2 Ex-vivo Imaging Dataset
		2.1 Specimen Preparation and Imaging
		2.2 Quantitative NFT Burden Maps from Histology
		2.3 Histology-Guided MTL Subregion Segmentations
	3 Methods
		3.1 Overview of Topological Unfolding Framework
		3.2 Segmentation of the Outer MTL Boundary in Ex vivo MRI
		3.3 Laplacian Coordinate System
		3.4 Mapping Image and Morphological Features to Unfolded Space
	4 Experiments and Results
		4.1 Consensus MTL Subregion Segmentation in Unfolded Coordinate Space
		4.2 Correlating NFT Burden with MTL Neurodegeneration
		4.3 Surface-Based Registration Using Mean Curvature Maps
	5 Conclusions
	References
Distinguishing Healthy Ageing from Dementia: A Biomechanical Simulation of Brain Atrophy Using Deep Networks
	1 Introduction
	2 Methods
		2.1 Data
		2.2 Preprocessing
		2.3 Model Overview
		2.4 Training and Evaluation
	3 Experimental Methods and Results
		3.1 Evaluation of Biomechanical Model
		3.2 Evaluation of Atrophy Estimation
	4 Discussion and Future Work
	References
Towards Self-explainable Classifiers and Regressors in Neuroimaging with Normalizing Flows
	1 Introduction
	2 Normalizing Flows as Generative Invertible Classifiers and Regressors
		2.1 Manifold-Constrained NFs for Efficient 3D Data Processing
		2.2 Implementation Details and Model Training
	3 Explainable AI with Normalizing Flows
		3.1 Derivative-Based Attribution Map of the Inverse
		3.2 Counterfactual Images for Systematic Analyses
	4 Experiments and Results
	5 Conclusion
	References
Patch vs. Global Image-Based Unsupervised Anomaly Detection in MR Brain Scans of Early Parkinsonian Patients
	1 Introduction
	2 Brain Anomaly Detection Pipeline
		2.1 Autoencoder Architectures
		2.2 Post-processing of the Reconstruction Error Maps
	3 Experiments
		3.1 Data
		3.2 Training of the Auto-Encoders
		3.3 Performance Evaluation
	4 Results
	5 Discussion and Conclusion
	References
MRI Image Registration Considerably Improves CNN-Based Disease Classification
	1 Introduction
	2 Data and Methods
		2.1 Dataset
		2.2 Image Preprocessing
		2.3 Network Architecture and Training
	3 Results
	4 Discussion
	References
Dynamic Sub-graph Learning for Patch-Based Cortical Folding Classification
	1 Introduction
	2 Methods
	3 Experimental Results
	4 Conclusions
	References
Detection of Abnormal Folding Patterns with Unsupervised Deep Generative Models
	1 Introduction
	2 Methods
		2.1 Focusing on Folding Information
		2.2 Generating Synthetic Brain Anomalies
		2.3 Learning a Representation of the Normal Variability
	3 Results
		3.1 Datasets and Implementation
		3.2 Analysing Learned Folding Variability
	4 Discussion and Conclusion
	References
PialNN: A Fast Deep Learning Framework for Cortical Pial Surface Reconstruction
	1 Introduction
	2 Related Work
	3 Method
		3.1 Deformation Block
		3.2 Smoothing and Training
	4 Experiments
	5 Conclusion
	References
Multi-modal Brain Segmentation Using Hyper-Fused Convolutional Neural Network
	1 Introduction
	2 Method
		2.1 Baseline Architecture
		2.2 Proposed Architecture
		2.3 Learning Process and Implementation Details
	3 Experiments and Results
		3.1 Datasets
		3.2 Results and Discussion
	4 Conclusion
	References
Robust Hydrocephalus Brain Segmentation via Globally and Locally Spatial Guidance
	1 Introduction
	2 Method
		2.1 Guidance with Registration Module
		2.2 Segmentation with Positional Correlation Attention Block
		2.3 Training Strategy
	3 Experiments and Results
		3.1 Datasets and Experiments
		3.2 Results
	4 Conclusion
	References
Brain Networks and Time Series
Geometric Deep Learning of the Human Connectome Project Multimodal Cortical Parcellation
	1 Introduction
	2 Methods
		2.1 Participants and Image Acquisition
		2.2 Modelling the Cortex as an Icosphere
		2.3 Image Processing and Augmentation
		2.4 Model Architecture and Implementation
	3 Results
	4 Discussion
	References
Deep Stacking Networks for Conditional Nonlinear Granger Causal Modeling of fMRI Data
	1 Introduction
	2 Materials and Methods
		2.1 Deep Stacking Network
		2.2 Conditional Nonlinear Granger Causal Modeling with DSN
		2.3 Model Validation and Application
	3 Experiments and Results
		3.1 Synthetic Dataset
		3.2 Simulated fMRI Dataset
		3.3 Real-World fMRI Dataset
	4 Discussion
	5 Conclusion
	References
Dynamic Adaptive Spatio-Temporal Graph Convolution for fMRI Modelling
	1 Introduction
	2 Methodology
		2.1 Preliminaries
		2.2 Temporal Lag Correction
		2.3 Temporal Feature Extraction
		2.4 Spatial Feature Extraction
		2.5 Framework of the Model
	3 Experiments
		3.1 Dataset
		3.2 Experimental Setup
		3.3 Experimental Results
	4 Generalizability
	5 Limitations
	6 Discussion
	References
Structure-Function Mapping via Graph Neural Networks
	1 Introduction
	2 Preliminaries
		2.1 Problem Statement
		2.2 Autoencoder
		2.3 Graph Convolutional Networks (GCN)
		2.4 Graph Transformer Networks (GTN)
	3 Experiments
		3.1 Data
		3.2 Implementation
	4 Results and Discussion
	5 Conclusion
	References
Improving Phenotype Prediction Using Long-Range Spatio-Temporal Dynamics of Functional Connectivity
	1 Introduction
	2 Related Works
	3 Methods
	4 Results
	5 Discussion
	References
H3K27M Mutations Prediction for Brainstem Gliomas Based on Diffusion Radiomics Learning
	1 Introduction
	2 Proposed Method
		2.1 Materials and Preprocessing
		2.2 Problem Formulation
		2.3 Multi-Mechanism Diffusion-Convolution (MMDC)
		2.4 Diffusion-Radiomics
	3 Experiments and Results
		3.1 Experimental Settings
		3.2 H3K27M Mutation Prediction Results
		3.3 Node Pooling Interpretation
	4 Conclusion
	References
Constrained Learning of Task-Related and Spatially-Coherent Dictionaries from Task fMRI Data
	1 Introduction
	2 Constrained Online Dictionary Learning
		2.1 Dictionary Learning of fMRI Data
		2.2 Incorporating Task Characteristics
		2.3 Constraining Spatial Patterns
		2.4 Optimization
	3 Application and Results
		3.1 Synthetic fMRI Data Generation Using SimTB
		3.2 Evaluation of Sparse Dictionary Learning Algorithms
		3.3 Synthetic Data Results
		3.4 Real Task fMRI Data
	4 Conclusion
	References
Author Index