Preface
Organization
Contents – Part VIII
Clinical Applications – Neuroimaging
CoactSeg: Learning from Heterogeneous Data for New Multiple Sclerosis Lesion Segmentation
1 Introduction
2 Datasets
3 Method
3.1 Overview
3.2 Multi-head Architecture
3.3 Longitudinal Relation Regularization
4 Results
5 Conclusion
References
Generating Realistic Brain MRIs via a Conditional Diffusion Probabilistic Model
1 Introduction
2 Methodology
2.1 Diffusion Probabilistic Model
2.2 Conditional Generation with DPM (cDPM)
2.3 Network Architecture
3 Experiments
3.1 Data
3.2 Implementation Details
3.3 Quantitative Comparison
3.4 Results
4 Conclusion
References
Towards Accurate Microstructure Estimation via 3D Hybrid Graph Transformer
1 Introduction
2 3D Hybrid Graph Transformer
2.1 Network Overview
2.2 Efficient q-Space Learning Module
2.3 3D x-Space Learning Module
3 Experiments
3.1 Implementation Details
3.2 Dataset and Evaluation Metrics
3.3 Experimental Results
3.4 Ablation Study
4 Conclusion
References
Cortical Analysis of Heterogeneous Clinical Brain MRI Scans for Large-Scale Neuroimaging Studies
1 Introduction
2 Methods
2.1 Learning of SDFs
2.2 Geometry Processing for Surface Placement
2.3 Implementation Details
3 Experiments and Results
3.1 Datasets
3.2 Competing Methods
3.3 Results on the ADNI Dataset
3.4 Results on the Clinical Dataset
3.5 Discussion and Conclusion
References
Flow-Based Geometric Interpolation of Fiber Orientation Distribution Functions
1 Introduction
2 Method
2.1 FOD Decomposition
2.2 Modeling Single Peak FOD Components as Flow of Vector Fields
2.3 Rotation Calculation for SPHARM-Based FODs
2.4 Evaluation Methods
3 Experiment Results
4 Conclusion
References
Learnable Subdivision Graph Neural Network for Functional Brain Network Analysis and Interpretable Cognitive Disorder Diagnosis
1 Introduction
2 Method
2.1 Preliminary
2.2 Functional Subdivision Block
2.3 Functional Aggregation Block
2.4 Objective Function
3 Experiments
3.1 Dataset and Experimental Settings
3.2 Result Analysis
3.3 Ablation Study
3.4 Interpretability of Brain States
4 Conclusion
References
FE-STGNN: Spatio-Temporal Graph Neural Network with Functional and Effective Connectivity Fusion for MCI Diagnosis
1 Introduction
2 Method
2.1 Local Spatial Structural Features and Short-Term Temporal Characteristics Extraction
2.2 Spatio-Temporal Fusion with Dynamic FC and EC
3 Experiments
3.1 Dataset and Experimental Settings
3.2 Ablation Studies
3.3 Comparison with Other Methods
4 Conclusion
References
Learning Normal Asymmetry Representations for Homologous Brain Structures
1 Introduction
2 Methods
2.1 Pre-training the Shape Characterization Encoder as a CAE
2.2 Learning Normal Asymmetries with a Siamese Network
3 Experimental Setup
4 Results and Discussion
4.1 Characterization of Normal and Disease Related Asymmetries
4.2 Comparison with Other Approaches
5 Conclusions
References
Simulation of Arbitrary Level Contrast Dose in MRI Using an Iterative Global Transformer Model
1 Introduction
2 Methods
3 Experiments and Results
4 Discussions and Conclusion
References
Development and Fast Transferring of General Connectivity-Based Diagnosis Model to New Brain Disorders with Adaptive Graph Meta-Learner
1 Introduction
2 Methods
2.1 Notation and Problem Formulation
2.2 Meta-Learner Training Algorithm
2.3 Multi-view Graph Classifier c
2.4 Meta-Controller m
3 Experiments
3.1 Dataset
3.2 Settings
3.3 Results and Discussions
4 Conclusion
References
Brain Anatomy-Guided MRI Analysis for Assessing Clinical Progression of Cognitive Impairment with Structural MRI
1 Introduction
2 Materials and Proposed Method
3 Experiment
4 Conclusion and Future Work
References
Dynamic Structural Brain Network Construction by Hierarchical Prototype Embedding GCN Using T1-MRI
1 Introduction
2 Methods
2.1 Backbone
2.2 Dynamic Hierarchical Prototype Learning
2.3 Brain Network Graph Construction and Classification
3 Experiments
3.1 Dataset
3.2 Implementation Details
4 Results
4.1 Comparing with SOTA Methods
4.2 Ablation Study
5 Conclusion
References
Microstructure Fingerprinting for Heterogeneously Oriented Tissue Microenvironments
1 Introduction
2 Methods
2.1 Fingerprint Dictionary
2.2 Solving the Bloch-Torrey Partial Differential Equation (BT-PDE)
2.3 Solving for Volume Fractions
2.4 Radius Bias Correction
3 Experiments
3.1 Volume Fraction
3.2 Cell Size and Membrane Permeability
3.3 In-vivo Data
3.4 Histological Corroboration
4 Conclusion
References
AUA-dE: An Adaptive Uncertainty Guided Attention for Diffusion MRI Models Estimation
1 Introduction
2 Method
2.1 q-t Space Sparsity
2.2 Adaptive Uncertainty Attention Modelling
2.3 Dataset and Training
3 Experiments and Results
3.1 Ablation Study
3.2 Performance Test
4 Conclusion
References
Relaxation-Diffusion Spectrum Imaging for Probing Tissue Microarchitecture
1 Introduction
2 Methods
2.1 Multi-compartment Model
2.2 Model Simplification via Spherical Mean
2.3 Estimation of Relaxation and Diffusion Parameters
2.4 Microstructure Indices
2.5 Data Acquisition and Processing
3 Results
3.1 Ex Vivo Data: Compartment-Specific Parameters
3.2 In Vivo Data: Compartment-Specific Parameters
3.3 In Vivo Data: Neurite Morphology
3.4 Relation Between Relaxation and Diffusivity
3.5 fODFs
4 Conclusion
References
Joint Representation of Functional and Structural Profiles for Identifying Common and Consistent 3-Hinge Gyral Folding Landmark
1 Introduction
2 Method
2.1 Dataset and Preprocessing
2.2 Joint Representation of Functional and Structural Profiles
2.3 Consistency Analysis from Anatomical, Structural and Functional Perspective
2.4 Comparative Analysis of Consistent 3-hinges for Structural Data and Multimodal Data
3 Result
3.1 Visualization of the Identified Consistent 3-hinges
3.2 Effectiveness of the Proposed Consistent 3-hinges
3.3 Comparative Analysis on the Consistent 3-hinges Based on Structural Data and Multimodal Data
4 Conclusion
References
Single-subject Multi-contrast MRI Super-resolution via Implicit Neural Representations
1 Introduction
2 Methods
3 Experiments and Results
4 Discussion and Conclusion
References
DeepSOZ: A Robust Deep Model for Joint Temporal and Spatial Seizure Onset Localization from Multichannel EEG Data
1 Introduction
2 Methodology
2.1 The DeepSOZ Model Architecture
2.2 Loss Function and Model Training
2.3 Model Validation
3 Experimental Results
4 Conclusion
References
Multimodal Brain Age Estimation Using Interpretable Adaptive Population-Graph Learning
1 Introduction
2 Methods
3 Experiments
3.1 Results
3.2 Ablation Studies
3.3 Interpretability
4 Conclusion
References
BrainUSL: Unsupervised Graph Structure Learning for Functional Brain Network Analysis
1 Introduction
2 Method
2.1 Graph Generation Module
2.2 Topology-Aware Encoder
2.3 Objective Functions
3 Experiments and Results
3.1 Dataset and Experimental Details
3.2 Classification Results
3.3 Functional Connectivity Analysis
3.4 Association of Brain Diseases
4 Conclusion
References
Learning Asynchronous Common and Individual Functional Brain Network for AD Diagnosis
1 Introduction
2 Proposed Method
2.1 Attention-Based Sparse Common-and-Individual FBN Construction Module (ASCFCM)
2.2 Cross Spatiotemporal Asynchronous FCs
3 Experiments
3.1 Data and Preprocessing
3.2 Experimental Settings
3.3 Experimental Results
3.4 Ablation Study
4 Visualization and Conclusion
References
Self-pruning Graph Neural Network for Predicting Inflammatory Disease Activity in Multiple Sclerosis from Brain MR Images
1 Introduction
2 Methodology
3 Experiments
4 Results
5 Conclusion
References
.28em plus .1em minus .1ematTRACTive: Semi-automatic White Matter Tract Segmentation Using Active Learning
1 Introduction
2 Methods
2.1 Binary Classification for Tract Segmentation
2.2 Active Learning for Tract Selection
3 Experiments
3.1 Data
3.2 Experimental Setup
3.3 Results
4 Discussion
References
Domain-Agnostic Segmentation of Thalamic Nuclei from Joint Structural and Diffusion MRI
1 Introduction
2 Methods
2.1 Training Dataset, Preprocessing, and Data Representation
2.2 Domain Randomisation and Data Augmentation
2.3 Loss
2.4 Architecture and Implementation Details
3 Experiments and Results
3.1 MRI Data
3.2 Competing Methods and Ablations
3.3 Results
4 Discussion and Conclusion
References
Neural Pre-processing: A Learning Framework for End-to-End Brain MRI Pre-processing
1 Introduction
2 Methods
2.1 Model
2.2 Loss Function
3 Experiments
3.1 Runtime Analyses
3.2 Pre-processing Performance
3.3 Ablation
4 Conclusion
References
Dynamic Functional Connectome Harmonics
1 Introduction
2 Methods
3 Results
4 Discussion
5 Conclusion
References
Wasserstein Distance-Preserving Vector Space of Persistent Homology
1 Introduction
2 Wasserstein Distance-Preserving Vector Space
2.1 One-Dimensional Persistence Diagrams
2.2 Closed-Form Wasserstein Distance for Different-Size Networks
2.3 Vector Representation of Persistence Diagrams
3 Application to Functional Brain Networks
References
Community-Aware Transformer for Autism Prediction in fMRI Connectome
1 Introduction
2 Method
2.1 Overview
2.2 Local-Global Transformer Encoder
2.3 Graph Readout Layer
3 Experiments
3.1 Datasets and Experimental Settings
3.2 Quantitative and Qualitative Results
3.3 Ablation Studies
4 Conclusion
References
Overall Survival Time Prediction of Glioblastoma on Preoperative MRI Using Lesion Network Mapping
1 Introduction
2 Materials and Methods
2.1 Materials
2.2 Methods
3 Experiments and Results
3.1 Experimental Settings
3.2 Comparison Studies
3.3 Brain Regions in Relation to GBM Survival
4 Conclusion
References
Exploring Brain Function-Structure Connectome Skeleton via Self-supervised Graph-Transformer Approach
1 Introduction
2 Method
2.1 Overview
2.2 Data and Preprocess
2.3 TSGR Framework
2.4 Analyzing Brain Key Connectome ROIs and Hierarchical Networks
2.5 Exploring the Connectome Skeleton of Brain ROIs
3 Experiments and Results
3.1 Experimental Performance
3.2 Analysis of Key Brain ROIs and Network Hierarchy
3.3 Analysis of Connectome Skeleton in Brain Networks
4 Conclusion
References
Vertex Correspondence in Cortical Surface Reconstruction
1 Introduction
2 Methods
3 Experiments and Results
4 Conclusion
References
Path-Based Heterogeneous Brain Transformer Network for Resting-State Functional Connectivity Analysis
1 Introduction
2 Methodology
2.1 Path-Based Heterogeneous Graph Generation
2.2 HP-GTC: Heterogeneous Path Graph Transformer Convolution Module to Learn Compact Features
2.3 Readout and Prediction
3 Experimental Results
4 Conclusion
References
Dynamic Graph Neural Representation Based Multi-modal Fusion Model for Cognitive Outcome Prediction in Stroke Cases
1 Introduction
2 Methodology
2.1 Graph Construction and Node Feature Extraction
2.2 Missing Information Compensation Module
2.3 Dynamic Graph Neural Representation
3 Materials and Experiments
3.1 Data and Preparation
3.2 Evaluation Measures
3.3 Experimental Design
4 Results
4.1 Effectiveness of Missing Information Compensation
4.2 Importance of Graph-Based Analysis
4.3 Superiority of Proposed Fusion Model
5 Conclusion and Discussion
References
Predicting Diverse Functional Connectivity from Structural Connectivity Based on Multi-contexts Discriminator GAN
1 Introduction
2 Method
2.1 Model Overview
2.2 MCGAN
3 Experiments
3.1 Setup
3.2 Results
4 Discussion
5 Conclusion
References
DeepGraphDMD: Interpretable Spatio-Temporal Decomposition of Non-linear Functional Brain Network Dynamics
1 Introduction
2 Methodology
2.1 Graph Dynamic Mode Decomposition
2.2 Adaptation of Graph-DMD for Nonlinear Graph Dynamics
2.3 Window-Based GraphDMD
3 Experiments
3.1 Dataset
3.2 Baseline Methods
3.3 Simulation Study
3.4 Application of GraphDMD and DeepGraphDMD in HCP Data
4 Results
4.1 Simulation Study
4.2 Application of GraphDMD and DeepGraphDMD in HCP Data
5 Conclusion
References
Disentangling Site Effects with Cycle-Consistent Adversarial Autoencoder for Multi-site Cortical Data Harmonization
1 Introduction
2 Method
2.1 Vanilla Autoencoder (AE)
2.2 Disentangled Autoencoder (DAE)
2.3 Cycle-Consistent Disentangled Autoencoder (CDAE)
3 Experiments and Results
3.1 Experimental Setting
3.2 Results
4 Conclusion
References
SurfFlow: A Flow-Based Approach for Rapid and Accurate Cortical Surface Reconstruction from Infant Brain MRI
1 Introduction
2 Methods
2.1 Overview
2.2 Dual-Modal Input
2.3 Loss Function
2.4 Deformation Computation in DMD Modules
2.5 Implementation Details
3 Results
3.1 Data
3.2 Evaluation Metrics
3.3 Results
3.4 Ablation Study
4 Conclusion
References
Prior-Driven Dynamic Brain Networks for Multi-modal Emotion Recognition
1 Introduction
2 Method
3 Experiment Results
4 Conclusion
References
Unified Surface and Volumetric Inference on Functional Imaging Data
1 Introduction
2 Methods
3 Results
4 Discussion
References
TractCloud: Registration-Free Tractography Parcellation with a Novel Local-Global Streamline Point Cloud Representation
1 Introduction
2 Methods
2.1 Training and Testing Datasets
2.2 TractCloud Framework
2.3 Implementation Details
3 Experiments and Results
3.1 Performance on the Labeled Atlas Dataset
3.2 Performance on the Independently Acquired Testing Datasets
4 Discussion and Conclusion
References
Robust and Generalisable Segmentation of Subtle Epilepsy-Causing Lesions: A Graph Convolutional Approach
1 Introduction
2 Methods
2.1 Graph Convolutional Network (GCN) for Surface-Based Lesion Segmentation
2.2 Data Augmentation
3 Experiments and Results
3.1 Dataset and Implementation Details
3.2 Results
4 Conclusions and Future Work
References
Weakly Supervised Cerebellar Cortical Surface Parcellation with Self-Visual Representation Learning
1 Introduction
2 Method
2.1 Cerebellar Surface Reconstruction and Geometric Feature Computation
2.2 Weakly Supervised Cerebellar Patch Representation Learning
2.3 Mapping from Latent Space to Parcellation Labels
3 Experiments
3.1 Dataset and Implementation
3.2 Comparison with the State-of-the-Art Methods
3.3 Different Features’ Influence Analysis
4 Conclusion
References
Maximum-Entropy Estimation of Joint Relaxation-Diffusion Distribution Using Multi-TE Diffusion MRI
1 Introduction
2 Method
2.1 On the Hausdorff Moment Problem
2.2 Maximum-Entropy Estimation
2.3 Dual Energy Minimization Problems
3 Examples
3.1 Synthetic Data
3.2 Comparison Methods
3.3 In Vivo rdMRI
4 Results
5 Summary
References
Physics-Informed Conditional Autoencoder Approach for Robust Metabolic CEST MRI at 7T
1 Introduction
2 Methods
3 Results
4 Discussion
5 Conclusion
References
A Coupled-Mechanisms Modelling Framework for Neurodegeneration
1 Introduction
2 Methodology
2.1 Model Definition
2.2 Bayesian Framework
2.3 Variational Inference
3 Experiments and Results
3.1 Data Processing
3.2 Results
4 Conclusions
References
A Texture Neural Network to Predict the Abnormal Brachial Plexus from Routine Magnetic Resonance Imaging
1 Introduction
2 Materials and Method
2.1 Dataset Preparation and Preprocessing
2.2 Triple Point Pattern (TPP)
2.3 TPPNet
3 Experiments
3.1 Preparations
3.2 Ablation Studies
3.3 Comparisons
4 Conclusions
References
Microscopy
Mitosis Detection from Partial Annotation by Dataset Generation via Frame-Order Flipping
1 Introduction
2 Method: Mitosis Detection with Partial Labels
2.1 Labeled Dataset Generation
2.2 Mitosis Detection with Generated Dataset
3 Experiments
4 Conclusion
References
CircleFormer: Circular Nuclei Detection in Whole Slide Images with Circle Queries and Attention
1 Introduction
2 Method
2.1 Overview
2.2 Representing Query with Anchor Circle
2.3 Circle Cross Attention
2.4 Circle Regression
2.5 Circle Instance Segmentation
2.6 Generalized Circle IoU
3 Experiment
3.1 Dataset and Evaluation
3.2 Implementation Details
3.3 Main Results
3.4 Ablation Studies
4 Conclusion
References
A Motion Transformer for Single Particle Tracking in Fluorescence Microscopy Images
1 Introduction
2 Method
2.1 Hypothesis Tree Construction
2.2 MoTT Network
2.3 Modeling Discrete Optimization Problem
2.4 Track Management
3 Experimental Results
3.1 Quantitative Performance
3.2 Robustness Analysis
4 Conclusion
References
B-Cos Aligned Transformers Learn Human-Interpretable Features
1 Introduction
2 Related Work
3 Methods
4 Implementation and Evaluation Details
5 Results and Discussion
6 Generalization to Other Architectures
7 Conclusion
References
PMC-CLIP: Contrastive Language-Image Pre-training Using Biomedical Documents
1 Introduction
2 The PMC-OA Dataset
2.1 Dataset Collection
2.2 Dataset Overview
3 Visual-language Pre-training
4 Experiment Settings
4.1 Pre-training Datasets
4.2 Downstream Tasks
4.3 Implementation Details
5 Result
5.1 PMC-OA surpasses SOTA large-scale biomedical dataset
5.2 PMC-CLIP achieves SOTA across downstream tasks
5.3 Ablation Study
6 Conclusion
References
Self-supervised Dense Representation Learning for Live-Cell Microscopy with Time Arrow Prediction
1 Introduction
2 Method
3 Experiments
3.1 Datasets
3.2 Implementation Details:
3.3 Time Arrow Prediction Pretraining
3.4 Downstream Tasks
4 Discussion
References
Learning Large Margin Sparse Embeddings for Open Set Medical Diagnosis
1 Introduction and Related Work
2 Method
2.1 Preliminaries
2.2 Margin Loss with Adaptive Scale (MLAS)
2.3 Open-Space Suppression (OSS)
2.4 Open Margin Cosine Loss (OMCL)
3 Result
3.1 Datasets, Evaluation Metrics, and Implementation Details
3.2 Comparison with State-of-the-Art Methods
3.3 Ablation Studies
4 Conclusion
References
Exploring Unsupervised Cell Recognition with Prior Self-activation Maps
1 Introduction
2 Method
2.1 Proxy Task
2.2 Prior Self-activation Map
2.3 Downstream Tasks
3 Experiments
3.1 Implementation Details
3.2 Result
4 Conclusion
References
Prompt-Based Grouping Transformer for Nucleus Detection and Classification
1 Introduction
2 Methodology
2.1 Transformer-Based Centroid Detector
2.2 Grouping Transformer Based Classifier
2.3 Loss Function
2.4 Grouping Prompts Based Tuning
3 Experiments and Results
3.1 Datasets and Implementation Details
3.2 Comparison with the State-of-the-Art
3.3 Ablation Analysis
4 Conclusion
References
PAS-Net: Rapid Prediction of Antibiotic Susceptibility from Fluorescence Images of Bacterial Cells Using Parallel Dual-Branch Network
1 Introduction
1.1 Method
1.2 Feature Interaction Unit
1.3 Hierarchical Multi-head Self-attention
2 Experiments and Results
2.1 Experimental Setup
2.2 Results
2.3 Robustness to HEp-2 Dataset
3 Conclusion
References
Diffusion-Based Data Augmentation for Nuclei Image Segmentation
1 Introduction
2 Method
2.1 Unconditional Nuclei Structure Synthesis
2.2 Conditional Histopathology Image Synthesis
3 Experiments and Results
3.1 Implementation Details
3.2 Effectiveness of the Proposed Data Augmentation Method
4 Conclusion
References
Unsupervised Learning for Feature Extraction and Temporal Alignment of 3D+t Point Clouds of Zebrafish Embryos
1 Introduction
2 Methods
2.1 Autoencoder
2.2 Regression Network
3 Experimental Results and Discussions
3.1 Data Sets and Evaluation
3.2 Experimental Settings
3.3 Experimental Results
4 Conclusion
References
3D Mitochondria Instance Segmentation with Spatio-Temporal Transformers
1 Introduction
2 Related Work
3 Method
3.1 Baseline Framework
3.2 Spatio-Temporal Transformer Res-UNET (STT-UNET)
4 Experiments
4.1 Results
5 Conclusion
References
Prompt-MIL: Boosting Multi-instance Learning Schemes via Task-Specific Prompt Tuning
1 Introduction
2 Method
2.1 Visual Prompt Tuning
2.2 Optimization
3 Experiments and Discussion
3.1 Datasets
3.2 Implementation Details
3.3 Results
4 Conclusion
References
Pick and Trace: Instance Segmentation for Filamentous Objects with a Recurrent Neural Network
1 Introduction
2 Method
2.1 Pick: Tip Points Detection Module
2.2 Trace: A Recurrent Network for Filament Tracing
3 Experiments
3.1 Synthetic Dataset
3.2 Microtubule Dataset
3.3 P. Rubescens Dataset
3.4 C. Elegans Dataset
4 Conclusion
References
BigFUSE: Global Context-Aware Image Fusion in Dual-View Light-Sheet Fluorescence Microscopy with Image Formation Prior
1 Introduction
2 Methods
2.1 Revisiting Dual-View LSFM Fusion Using Bayes
2.2 Image Clarity Characterization with Image Formation Prior
2.3 Least Squares Smoothness of Focus-Defocus Boundary
2.4 Focus-Defocus Boundary Inference via EM
2.5 Competitive Methods
3 Results and Discussion
3.1 Evaluation on LSFM Images with Synthetic Blur
3.2 Evaluation on LSFM Images with Real Blur
4 Conclusion
References
LUCYD: A Feature-Driven Richardson-Lucy Deconvolution Network
1 Introduction
2 Method
2.1 Correction Module and Bottleneck
2.2 Update Module
2.3 Loss Function
3 Experiments
3.1 Setup
3.2 Results
4 Conclusion
References
Deep Unsupervised Clustering for Conditional Identification of Subgroups Within a Digital Pathology Image Set
1 Introduction
2 Methods
2.1 Variational Deep Embedding (VaDE)
2.2 Conditionally Decoded Variational Deep Embedding (CDVaDE)
2.3 Deep Embedding Clustering (DEC)
2.4 Related Works in Medical Imaging
3 Experiments
3.1 Colored MNIST
3.2 Application to a Digital Pathology Dataset
4 Conclusion
References
Weakly-Supervised Drug Efficiency Estimation with Confidence Score: Application to COVID-19 Drug Discovery
1 Introduction
2 Dataset
3 Methods
3.1 Image Preprocessing and Embedding
3.2 Data Augmentation with Weak Labels
3.3 Confident Hit Predictor
4 Experiments and Results
4.1 Experimental Settings
4.2 Representation Quality Assurance
4.3 Disease Scores Quality Assurance
4.4 Confidence Scores Quality Assurance
4.5 Evaluation
5 Conclusion
References
Author Index