FunduSAM: A Specialized Deep Learning Model for Enhanced Optic Disc and Cup Segmentation in Fundus Images

2024-12-03

IEEE International Conference on Bioinformatics and Biomedicine

AI Summary

Existing SAM-based methods exhibit limited performance in optic disc (OD) and optic cup (OC) segmentation from fundus images due to low contrast, ill-defined boundaries, and complex anatomical structures. To address these challenges, we propose FunduSAM—the first dedicated segmentation framework integrating adapter-based parameter-efficient fine-tuning, CBAM-based channel-spatial joint attention, and polar coordinate transformation—designed to preserve OD/OC anatomical consistency and enhance boundary localization accuracy. Built upon the SAM architecture, FunduSAM employs a composite loss function to improve segmentation robustness. Evaluated on the REFUGE dataset (1,200 images), FunduSAM outperforms five state-of-the-art methods, achieving new SOTA results in both Dice coefficient and Hausdorff distance for OD and OC segmentation. These results demonstrate its superior structural awareness and clinical applicability.

Technology Category

• Computer Vision: Segmentation
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Data transparency and provenance
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The Segment Anything Model (SAM) has gained popularity as a versatile image segmentation method, thanks to its strong generalization capabilities across various domains. However, when applied to optic disc (OD) and optic cup (OC) segmentation tasks, SAM encounters challenges due to the complex structures, low contrast, and blurred boundaries typical of fundus images, leading to suboptimal performance. To over-come these challenges, we introduce a novel model, FunduSAM, which incorporates several Adapters into SAM to create a deep network specifically designed for OD and OC segmentation. The FunduSAM utilizes Adapter into each transformer block after encoder for parameter fine-tuning (PEFT). It enhances SAM’s feature extraction capabilities by designing a Convolutional Block Attention Module (CBAM), addressing issues related to blurred boundaries and low contrast. Given the unique requirements of OD and OC segmentation, polar transformation is used to convert the original fundus OD images into a format better suited for training and evaluating FunduSAM. A joint loss is used to achieve structure preservation between the OD and OC, while accurate segmentation. Extensive experiments on the REFUGE dataset, comprising 1,200 fundus images, demonstrate the superior performance of FunduSAM compared to five mainstream approaches.

Problem

• Enhance optic disc and cup segmentation
• Overcome low contrast and blurred boundaries
• Develop specialized deep learning model FunduSAM

Innovation

• Adapter for parameter fine-tuning
• Convolutional Block Attention Module
• Polar transformation for image format