Interpretability of Machine Intelligence in Medical Image Computing, and Topological Data Analysis and Its Applications for Medical Data (Image ... Vision, Pattern Recognition, and Graphics)

✍ Scribed by Mauricio Reyes (editor), Pedro Henriques Abreu (editor), Jaime Cardoso (editor), Mustafa Hajij (editor), Ghada Zamzmi (editor), Paul Rahul (editor), Lokendra Thakur (editor)

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 138
• Category: Library

✦ Synopsis

This book constitutes the refereed joint proceedings of the 4th International Workshop on Interpretability of Machine Intelligence in Medical Image Computing, iMIMIC 2020, and the First International Workshop on Topological Data Analysis and Its Applications for Medical Data, TDA4MedicalData 2021, held on September 27, 2021, in conjunction with the 24th International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2021.

The 7 full papers presented at iMIMIC 2021 and 5 full papers held at TDA4MedicalData 2021 were carefully reviewed and selected from 12 submissions each. The iMIMIC papers focus on introducing the challenges and opportunities related to the topic of interpretability of machine learning systems in the context of medical imaging and computer assisted intervention. TDA4MedicalData is focusing on using TDA techniques to enhance the performance, generalizability, efficiency, and explainability of the current methods applied to medical data.

✦ Table of Contents

iMIMIC 2021
Organization
TDA4MedicalData 2021
Organization
Contents
iMIMIC 2021 Workshop
Interpretable Deep Learning for Surgical Tool Management
1 Introduction
2 Surgical Tool Dataset Overview
2.1 Surgery Knowledge Base
3 Experimental Method
4 Results and Conclusions
References
Soft Attention Improves Skin Cancer Classification Performance
1 Introduction
2 Method
2.1 Dataset
2.2 Soft Attention
2.3 Model Setup
2.4 Loss Function
3 Results
3.1 Ablation Analysis
3.2 Quantitative Analysis
3.3 Qualitative Analysis
4 Conclusion
References
Deep Grading Based on Collective Artificial Intelligence for AD Diagnosis and Prognosis
1 Introduction
2 Materials and Method
2.1 Datasets
2.2 Preprocessing
2.3 Deep Grading for Disease Visualization
2.4 Collective AI for Grading
2.5 Graph Convolutional Neural Network for Classification
2.6 Implementation Details
3 Experimental Results
4 Conclusion
References
This Explains That: Congruent Image–Report Generation for Explainable Medical Image Analysis with Cyclic Generative Adversarial Networks
1 Introduction
1.1 Prior Work
1.2 Our Approach
1.3 Contributions
2 Method
2.1 Coherent Image-Report Pairs with CycleGANs
2.2 CycleGAN
2.3 Explanatory Image–Report Pairs
2.4 Dataset
2.5 Implementation
3 Evaluation
3.1 Evaluation of Generated Reports
3.2 Evaluation of Explanations
4 Conclusion
References
Visual Explanation by Unifying Adversarial Generation and Feature Importance Attributions
1 Introduction
2 Related Works
3 Method
3.1 Combining Gradient Attribution and Adversarial Methods
3.2 Choice of the Path and Regularization
4 Experiments and Results
4.1 Datasets and Models
4.2 Attribution Techniques and Implementation Details
4.3 Results
5 Conclusion
References
The Effect of the Loss on Generalization: Empirical Study on Synthetic Lung Nodule Data
1 Introduction
2 Materials and Methods
2.1 Data
2.2 Neural Network Architectures and Loss Functions
3 Experiments and Results
4 Discussion
5 Conclusion
References
Voxel-Level Importance Maps for Interpretable Brain Age Estimation
1 Introduction
2 Materials and Methods
2.1 Dataset and Preprocessing
2.2 Brain Age Estimation
2.3 Localisation
3 Results
3.1 Age Estimation
3.2 Population-Based Importance Maps
4 Discussion
5 Conclusion
References
TDA4MedicalData Workshop
Lattice Paths for Persistent Diagrams
1 Introduction
2 Methods
3 Application: Spike Proteins of COVID-19 Virus
4 Conclusions
References
Neighborhood Complex Based Machine Learning (NCML) Models for Drug Design
1 Introduction
2 Method
2.1 NC-Based Biomolecular Structure and Interaction Analysis
2.2 NC-Based Persistent Spectral Models
3 Experiments
3.1 Data
3.2 Model Settings
3.3 Results
4 Conclusions
References
Predictive Modelling of Highly Multiplexed Tumour Tissue Images by Graph Neural Networks
1 Introduction
2 Method
2.1 Building Graphs from Highly Multiplexed Imaging
2.2 Graph Neural Networks
3 Data Sets
4 Experiments and Results
4.1 Marker Selection and Preprocessing
4.2 Finding the Best Architecture
4.3 Comparing Standardization Strategies
5 Discussion
References
Statistical Modeling of Pulmonary Vasculatures with Topological Priors in CT Volumes
1 Introduction
2 Appearance Model of 3D Blood Vessels Based on -VAE
3 Topological Loss Function Based on Persistent Homology
3.1 Persistent Homology
3.2 Topological Loss Functions and Priors
4 Incorporating Topological Priors into Statistical Model
5 Experiments
5.1 Blood Vessels Containing Hole Artifacts
5.2 Blood Vessels with Bifurcations
6 Conclusion
References
Topological Detection of Alzheimer's Disease Using Betti Curves
1 Introduction
2 Related Works
3 Preliminaries
3.1 Homology and Betti Curves
3.2 Dataset
4 Algorithms and Methods
4.1 Betti Curve Based Features on MRI Data
4.2 Aging vs Alzheimer's
4.3 Implementation Details
5 Experimental Results
5.1 Brain Age Prediction
5.2 AD Detection with Age Correction
6 Discussion
References
Author Index