Wavelet-based Global-Local Interaction Network with Cross-Attention for Multi-View Diabetic Retinopathy Detection

2025-03-25

AI Summary

In multi-view diabetic retinopathy (DR) detection, challenges include large inter-lesion scale variations, spatially dispersed lesion distributions, and inadequate modeling of inter-view correlations and redundancies during cross-view feature fusion. To address these, we propose a novel dual-branch multi-view detection framework. Our method introduces a wavelet high-frequency branch to enhance lesion edge details, guided by global semantic information for precise local localization; additionally, we design a cross-view cross-attention module that explicitly models both complementarity and redundancy across views. The framework integrates wavelet decomposition, dual-branch CNNs, cross-view attention, and global–local interaction mechanisms. Evaluated on multiple mainstream public DR datasets, our approach achieves state-of-the-art performance, significantly improving detection sensitivity for micro-lesions and occluded lesions. The source code is publicly available.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Multi-view diabetic retinopathy (DR) detection has recently emerged as a promising method to address the issue of incomplete lesions faced by single-view DR. However, it is still challenging due to the variable sizes and scattered locations of lesions. Furthermore, existing multi-view DR methods typically merge multiple views without considering the correlations and redundancies of lesion information across them. Therefore, we propose a novel method to overcome the challenges of difficult lesion information learning and inadequate multi-view fusion. Specifically, we introduce a two-branch network to obtain both local lesion features and their global dependencies. The high-frequency component of the wavelet transform is used to exploit lesion edge information, which is then enhanced by global semantic to facilitate difficult lesion learning. Additionally, we present a cross-view fusion module to improve multi-view fusion and reduce redundancy. Experimental results on large public datasets demonstrate the effectiveness of our method. The code is open sourced on https://github.com/HuYongting/WGLIN.

Problem

• Detecting diabetic retinopathy across multiple views with incomplete lesions
• Addressing variable lesion sizes and scattered locations in multi-view DR
• Improving multi-view fusion by reducing redundancies and enhancing correlations

Innovation

• Wavelet-based two-branch network for lesion features
• Cross-view fusion module reduces information redundancy
• Global-local interaction enhances difficult lesion learning