This book constitutes the refereed proceedings of the 6th International Workshop on Ophthalmic Medical Image Analysis, OMIA 2020, held in conjunction with the 23rd International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2020, in Lima, Peru, in October 2020. The workshop was held virtually due to the COVID-19 crisis.
The 21 papers presented at OMIA 2020 were carefully reviewed and selected from 34 submissions. The papers cover various topics in the field of ophthalmic medical image analysis and challenges in terms of reliability and validation, number and type of conditions considered, multi-modal analysis (e.g., fundus, optical coherence tomography, scanning laser ophthalmoscopy), novel imaging technologies, and the effective transfer of advanced computer vision and machine learning technologies.
Author(s): Huazhu Fu, Mona K. Garvin, Tom MacGillivray, Yanwu Xu, Yalin Zheng
Series: Image Processing, Computer Vision, Pattern Recognition, and Graphics, 12069
Publisher: Springer
Year: 2020
Language: English
Pages: 218
City: Cham
Preface
Organization
Contents
Bio-inspired Attentive Segmentation of Retinal OCT Imaging
1 Introduction
2 Methods
2.1 Bio-inspired Attentive Segmentation
2.2 Low-Rank Oriented Attention (LROA)
2.3 Architectural Overview
3 Experiments and Results
3.1 Data
3.2 Experimental Setup
3.3 Results
4 Discussion and Conclusion
References
DR Detection Using Optical Coherence Tomography Angiography (OCTA):pg A Transfer Learning Approach with Robustness Analysis
1 Introduction
2 Methods
2.1 Datasets and Imaging Devices
2.2 Data Augmentation and Transfer Learning
3 Results
3.1 Classification of Controls, DR and NoDR Patients
3.2 Model Validation on OCTAGON Dataset
4 Discussion and Conclusions
References
What is the Optimal Attribution Method for Explainable Ophthalmic Disease Classification?
1 Introduction
2 Related Studies
3 Methods
4 Analysis
4.1 Quantitative Analysis
4.2 Qualitative Analysis
5 Conclusion
References
DeSupGAN: Multi-scale Feature Averaging Generative Adversarial Network for Simultaneous De-blurring and Super-Resolution of Retinal Fundus Images
1 Introduction
2 Related Work
3 Methodology
3.1 DeSupGAN Structure
3.2 Loss Functions
4 Experiment
4.1 Dataset Generation
4.2 Training Details
4.3 Results
4.4 Ablation Studies
5 Conclusion
References
Encoder-Decoder Networks for Retinal Vessel Segmentation Using Large Multi-scale Patches
1 Introduction
2 Studying Patch Size and Model Architecture
2.1 Effective Patch Sizes
2.2 Efficient Architecture
3 Comparison with State-of-the-Art
3.1 Results
3.2 High-Resolution Fundus Images
3.3 Cross-Dataset Evaluation
4 Conclusion
References
Retinal Image Quality Assessment via Specific Structures Segmentation
1 Introduction
2 Database
3 Method
3.1 Segmentation Modules
3.2 Quality Assessment Module
3.3 Implementation Detail
4 Results
4.1 Comparative Studies
4.2 Ablation Studies
4.3 Computational Complexity
5 Conclusion
References
Cascaded Attention Guided Network for Retinal Vessel Segmentation
1 Introduction
2 Methodology
2.1 Cascaded Deep Learning Network
2.2 Attention UNet++
3 Experiments
3.1 Datasets
3.2 Implementation Details
3.3 Evaluation Methods
3.4 Results
3.5 Ablation Study
4 Conclusion
References
Self-supervised Denoising via Diffeomorphic Template Estimation: Application to Optical Coherence Tomography
1 Introduction
2 Methods
2.1 Problem Formulation
2.2 Registration
2.3 Denoising
3 Experiments
3.1 Dataset
3.2 Implementation Details
3.3 Evaluation Methods
3.4 Results
4 Discussion
References
Automated Detection of Diabetic Retinopathy from Smartphone Fundus Videos
1 Introduction
2 Materials and Methods
2.1 Data Acquisition and Annotation
2.2 Cropping Frames to the Lens
2.3 Selection of Informative Frames
2.4 Classification of Referable Diabetic Retinopathy
3 Results
3.1 Evaluation of Informative Frame Selection
3.2 Evaluation of Disease Detection
3.3 Computational Effort
4 Discussion and Conclusion
References
Optic Disc, Cup and Fovea Detection from Retinal Images Using U-Net++ with EfficientNet Encoder
1 Introduction
2 Methodology
2.1 Dataset
2.2 Proposed Method
3 Results and Discussion
3.1 Experimental Set-up
3.2 Results and Discussion
4 Conclusion
References
Multi-level Light U-Net and Atrous Spatial Pyramid Pooling for Optic Disc Segmentation on Fundus Image
1 Introduction
2 Methodology
2.1 Light U-Net
2.2 Atrous Convolution and Spatial Pyramid Pooling
3 Experiments
3.1 Dataset and Evaluation Criteria
3.2 Implementation Details
3.3 Comparison with State-of-the-Art
3.4 Ablation Study
4 Conclusions
References
An Interactive Approach to Region of Interest Selection in Cytologic Analysis of Uveal Melanoma Based on Unsupervised Clustering
1 Introduction
2 Method
2.1 Step-1 Clustering
2.2 Step-2 Clustering
2.3 Interactive Centroid Assignment and Refinement
3 Experiment
3.1 Experiment Setup
3.2 Ablation Study for Clustering Algorithm
3.3 Ablation Study for Interactive Refinement
4 Conclusion
References
Retinal OCT Denoising with Pseudo-Multimodal Fusion Network
1 Introduction
2 Methods
3 Experiments
3.1 Data Set
3.2 Experimental Design
4 Results
4.1 Visual Analysis
4.2 Quantitative Evaluation
5 Conclusion and Future Work
References
Deep-Learning-Based Estimation of 3D Optic-Nerve-Head Shape from 2D Color Fundus Photographs in Cases of Optic Disc Swelling
1 Introduction
2 Methods
2.1 Overview
2.2 Neural Network Architecture
2.3 Statistical Total Retinal Shape Models in 3D
3 Experimental Methods
4 Results
5 Discussion and Conclusion
References
Weakly Supervised Retinal Detachment Segmentation Using Deep Feature Propagation Learning in SD-OCT Images
1 Introduction
2 Methodology
2.1 Saliency Map Generation Based on Improved CAM
2.2 Soft Label Using Feature Propagation Learning
2.3 Lesion Segmentation with Strong Supervised Network
3 Experiments
3.1 Datasets and Evaluation Metrics
3.2 Comparison Experiments
4 Conclusion
References
A Framework for the Discovery of Retinal Biomarkers in Optical Coherence Tomography Angiography (OCTA)
1 Introduction
2 Methods
2.1 Vascular Graph Construction
2.2 Graph Simplification
2.3 Feature Extraction
2.4 Demographics and Statistical Analysis
3 Results
4 Discussion and Conclusions
References
An Automated Aggressive Posterior Retinopathy of Prematurity Diagnosis System by Squeeze and Excitation Hierarchical Bilinear Pooling Network
1 Introduction
2 Methodology
2.1 Hierarchical Bilinear Pooling Module
2.2 Squeeze and Excitation Module
2.3 Focal-Loss
3 Experiments
3.1 Data
3.2 Results
4 Conclusions
References
Weakly-Supervised Lesion-Aware and Consistency Regularization for Retinitis Pigmentosa Detection from Ultra-Widefield Images
1 Introduction
2 Methodology
2.1 Multi-scale Global Average Pooling
2.2 Consistency Regularization
3 Experiments
3.1 Dataset and Implementation
3.2 Lesion Attention Map Visualization
3.3 Classification Performance
4 Conclusion
References
A Conditional Generative Adversarial Network-Based Method for Eye Fundus Image Quality Enhancement
1 Introduction
2 Pix2Pix-Fundus Oculi Quality Enhancer (P2P-FOQE)
2.1 Pre-enhancement
2.2 Pix2Pix Enhancement
2.3 Post-enhancement
2.4 Training the P2P-FOQE Model
3 Experimental Evaluation
3.1 Dataset for Eye Fundus Image Quality Enhancement
3.2 Experimental Setup
3.3 Results and Discussion
4 Conclusions
References
Construction of Quantitative Indexes for Cataract Surgery Evaluation Based on Deep Learning
1 Introduction
2 Proposed Method
2.1 ResUnet for Pupil Segmentation
2.2 Pretrained ResNet for Keratome Localization
2.3 Constructing the Evaluation Indexes of Incision
3 Experiment Results
3.1 Dataset
3.2 Model Settings
3.3 Result and Discussion
4 Conclusion
References
Hybrid Deep Learning Gaussian Process for Diabetic Retinopathy Diagnosis and Uncertainty Quantification
1 Introduction
2 Related Work
3 Deep Learning Gaussian Process for Diabetic Retinopathy Diagnosis (DLGP-DR)
3.1 Feature Extraction - Inception-V3
3.2 Gaussian Processes
4 Experimental Evaluation
4.1 Datasets
4.2 Experimental Setup
4.3 EyePACS Results
4.4 Messidor-2 Results
4.5 Discussion
5 Conclusions
References
Author Index
