Graphs in Biomedical Image Analysis, and Overlapped Cell on Tissue Dataset for Histopathology: 5th MICCAI Workshop, GRAIL 2023 and 1st MICCAI ... (Lecture Notes in Computer Science, 14373)

This LNCS conference volume constitutes the proceedings of the MICCAI Workshop GRAIL 2023 and MICCAI Challenge OCELOT 2023, Held in Conjunction with MICCAI 2023, Vancouver, BC, Canada, September 23, and October 4, 2023. The 9 full papers (GRAIL 2023) and 6 full papers (OCELOT 2023) included in this volume were carefully reviewed and selected from GRAIL 14 (GRAIL 2023) and 6 (OCELOT 2023) submissions. The conference GRAIL 2023 a wide set of methods and application and OCELOT 2023 focuses on the cover a wide range of methods utilizing tissue information for better cell detection, in the sense of training strategy, model architecture, and especially how to model cell-tissue relationships. GRAIL 2023 Preface OCELOT 2023 Preface Organization Contents GRAIL 2023 SCOPE: Structural Continuity Preservation for Retinal Vessel Segmentation 1 Introduction 2 Methodology 2.1 Scope 3 Experiments and Results 3.1 Experimental Setup 3.2 Results 4 Conclusion References Extended Graph Assessment Metrics for Regression and Weighted Graphs 1 Introduction 2 Background and Related Work 2.1 Definition of Graphs 2.2 Homophily 2.3 Cross-Class Neighbourhood Similarity 3 Extended Graph Metrics 3.1 CCNS Distance 3.2 K-Hop Metrics 3.3 Metrics for Continuous Adjacency Matrices 3.4 Homophily for Regression 3.5 Metric Evaluation 4 Experiments and Results 4.1 Datasets 4.2 GNN Training 4.3 Results 5 Conclusion and Future Work A Further Information on Extended Graph Assessment Metrics A.1 K-Hop metrics A.2 Node-wise metrics B Experiments B.1 Synthetic dataset B.2 TADPOLE dataset B.3 UKBB dataset B.4 Baseline results References Multi-head Graph Convolutional Network for Structural Connectome Classification 1 Introduction 2 Methods 3 Experiments 3.1 Datasets 3.2 Pre-processing 3.3 Results and Discussion 4 Conclusion References Tertiary Lymphoid Structures Generation Through Graph-Based Diffusion 1 Introduction 2 Background 3 Diffusion Based Cell-Graph Generation of Tertiary Lymphoid Structures 4 Experiments 5 Conclusion A Examples of High and Low TLS Content Subgraph B Distributions of TLS Embeddings References Prior-RadGraphFormer: A Prior-Knowledge-Enhanced Transformer for Generating Radiology Graphs from X-Rays 1 Introduction 2 Method 2.1 RadGraphFormer 2.2 Prior Knowledge Integration 2.3 Training and Inference 3 Experiments 3.1 Datasets 3.2 Implementation Details 3.3 Evaluation 4 Results and Discussion 4.1 Main Results 4.2 Downstream Tasks 4.3 Ablation Studies 4.4 Limitations and Outlook 5 Conclusion References A Comparative Study of Population-Graph Construction Methods and Graph Neural Networks for Brain Age Regression 1 Introduction 2 Methods 3 Experiments 3.1 Dataset 3.2 Results 3.3 Visualizations 3.4 Implementation Details 4 Discussion and Conclusions References Self Supervised Multi-view Graph Representation Learning in Digital Pathology 1 Introduction 2 Methods 2.1 Graph Construction 2.2 Graph Representation Learning 3 Experiments 4 Results 5 Conclusion References Multi-level Graph Representations of Melanoma Whole Slide Images for Identifying Immune Subgroups 1 Introduction 2 Related Works 3 Dataset 4 Methods 4.1 Segmentation and Feature Extraction 4.2 Graph Construction 4.3 Model Architecture 4.4 Model Training 5 Results 6 Discussion and Conclusion References Heterogeneous Graphs Model Spatial Relationship Between Biological Entities for Breast Cancer Diagnosis 1 Introduction 2 Related Work 3 Methodology 3.1 Tissue and Cell Graph Extraction from Histology Images 3.2 Heterogeneous Graph Convolution 3.3 Adaptive Weighted Aggregation for Multi-level Feature Fusion 3.4 Cross-Attention Feature Fusion and Attentive Interaction Using Transformers 4 Experiments and Results 4.1 Clinical Datasets and Evaluation Methods 4.2 Experimental Setup 4.3 Comparison with the State-of-the-Art Methods 4.4 Ablation Studies 5 Conclusion References OCELOT 2023 SoftCTM: Cell Detection by Soft Instance Segmentation and Consideration of Cell-Tissue Interaction 1 Introduction 2 Related Works 3 Methods 3.1 Dataset 3.2 Model Architecture 3.3 Tissue Segmentation Training 3.4 Cell Detection Ground Truth Formats 3.5 Cell Detection Training 3.6 Cell Detection Postprocessing 3.7 Combined Cell-Tissue Model 4 Results 4.1 Main Findings 4.2 Organ-Wise Results 5 Conclusion A Segmentation Ground Truth Generation with NuClick B Postprocessing for the Hard IS Ground Truth Format C Ablation Study: Different Values for Soft IS Ground Truth D Organ-Wise Performance on OCELOT Validation Set References Detecting Cells in Histopathology Images with a ResNet Ensemble Model 1 Introduction 2 Model Architecture 3 Dataset Preparation 4 Training Procedure 5 Inference Pipeline 6 Results and Discussion References Enhancing Cell Detection via FC-HarDNet and Tissue Segmentation: OCELOT 2023 Challenge Approach 1 Introduction 2 Method 2.1 Preliminary 2.2 FC-HarDNet 2.3 Model Ensemble 2.4 Loss Function 3 Experiment 3.1 Setting 3.2 Dataset 3.3 Results 4 Discussion 5 Conclusion References Dense Prediction of Cell Centroids Using Tissue Context and Cell Refinement 1 Introduction 2 Methodology 2.1 Data 2.2 Tissue Context Algorithm 2.3 Cell Refinement Algorithm 3 Results 4 Discussion and Conclusion A Excluded Image IDs B Output Crop Margin References Enhancing Cell Detection in Histopathology Images: A ViT-Based U-Net Approach 1 Introduction 2 Related Work 2.1 Cell Detection and Segmentation 2.2 Large-scale Pre-training Model 2.3 Parameter-Efficient Fine-Tuning Strategies 3 Method 3.1 Preprocessing 3.2 Cell-Tissue-ViT 3.3 Loss Function 4 Implementation Details 4.1 Environment Settings 4.2 Training Protocol 5 Results and Discussion 5.1 Quantitative Results on Validation Set 5.2 Model Ensemble 6 Conclusion References Generating BlobCell Label from Weak Annotations for Precise Cell Segmentation 1 Introduction 2 Related Work 2.1 Generate Cell Segmentation Labels from Cell Center Locations 2.2 Blob Extraction 2.3 Tissue Injection Model 3 Method 3.1 BlobCell Label 3.2 BlobCell-Tissue Prediction Injection Models 3.3 Implementation Details 4 Experiment and Results 5 Conclusion References Author Index