Advanced Segmentation of Diabetic Retinopathy Lesions Using DeepLabv3+

2024-10-22

ACS/IEEE International Conference on Computer Systems and Applications

AI Summary

To address low segmentation accuracy, high annotation cost, and severe inter-class coupling among microaneurysms, hemorrhages, and hard/soft exudates in diabetic retinopathy (DR), this paper proposes a lesion-type-decoupled four-channel binary segmentation framework to eliminate multi-class semantic ambiguity. It integrates L-channel CLAHE enhancement in the LAB color space with lesion-specific lightweight data augmentation to improve robustness under limited training samples. Leveraging DeepLabv3+ as the backbone, the method enables end-to-end joint modeling and multi-scale feature fusion. Evaluated on the IDRID dataset, the approach achieves 99% segmentation accuracy, significantly enhancing lesion localization precision and clinical interpretability. This work establishes an efficient, interpretable paradigm for medical image segmentation tasks characterized by scarce annotated data and high labeling costs.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Security and Privacy: Data transparency and provenance

Abstract

To improve the segmentation of diabetic retinopathy lesions (microaneurysms, hemorrhages, exudates, and soft exudates), we implemented a binary segmentation method specific to each type of lesion. As post-segmentation, we combined the individual model outputs into a single image to better analyze the lesion types. This approach facilitated parameter optimization and improved accuracy, effectively overcoming challenges related to dataset limitations and annotation complexity. Specific preprocessing steps included cropping and applying contrast-limited adaptive histogram equalization to the L channel of the LAB image. Additionally, we employed targeted data augmentation techniques to further refine the model's efficacy. Our methodology utilized the DeepLabv3+ model, achieving a segmentation accuracy of 99 %. These findings highlight the efficacy of innovative strategies in advancing medical image analysis, particularly in the precise segmentation of diabetic retinopathy lesions. The IDRID dataset was utilized to validate and demonstrate the robustness of our approach.

Problem

• Improves segmentation of diabetic retinopathy lesions using DeepLabv3+
• Combines individual model outputs for better lesion analysis
• Addresses dataset limitations and annotation complexity challenges

Innovation

• DeepLabv3+ for precise lesion segmentation
• Contrast-limited adaptive histogram equalization preprocessing
• Targeted data augmentation enhances model efficacy