Related Experiment Video
Updated: Aug 6, 2026
08:53
The Arteriovenous (AV) Loop in a Small Animal Model to Study Angiogenesis and Vascularized Tissue Engineering
Published on: November 2, 2016
AV-NeXt: a topology-aware framework with interaction modeling and competitive gating for retinal artery/vein
1School of Information Science and Engineering, Dalian Polytechnic University, Qinggongyuan 1, Ganjingzi District, Liaoning, 116034, Dalian, China.
Medical & Biological Engineering & Computing
|July 25, 2026
Summary
AV-NeXt improves retinal artery and vein segmentation by modeling interactions at crossings. This topology-aware framework enhances diagnostic accuracy for cardiovascular and ophthalmic diseases.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Automated segmentation and classification of retinal arteries and veins (A/V) are crucial for diagnosing systemic cardiovascular and ophthalmic diseases.
- Current deep learning models struggle with A/V crossings, leading to segmentation errors and classification ambiguity.
Purpose of the Study:
- To introduce AV-NeXt, a novel topology-aware framework designed to enhance A/V crossing recognition.
- To improve the accuracy of automated retinal vascular segmentation and classification.
Main Methods:
- AV-NeXt integrates anatomically motivated artery-vein interaction modeling into a ConvNeXt backbone.
- Key components include an Interaction-conditioned Crossing Head and a Competitive Gating Mechanism to refine crossing predictions and reduce ambiguity.
Main Results:
- AV-NeXt achieved superior performance on the Fundus-AVSeg dataset, with mDice of 76.72% and mIoU of 63.64%.
- The framework demonstrated strong crossing-region recognition (Crossing IoU of 44.85%) and competitive zero-shot performance on the RITE dataset (Crossing IoU of 28.52%).
Conclusions:
- AV-NeXt effectively addresses limitations in modeling local artery-vein interactions at crossings.
- The proposed framework offers improved accuracy and reliability for automated retinal vascular analysis, aiding in early disease diagnosis.
