![[2602.15072] GRAFNet: Multiscale Retinal Processing via Guided Cortical Attention Feedback for Enhancing Medical Image Polyp Segmentation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15072] GRAFNet: Multiscale Retinal Processing via Guided Cortical Attention Feedback for Enhancing Medical Image Polyp Segmentation
Summary
GRAFNet introduces a novel architecture for polyp segmentation in colonoscopy, enhancing accuracy through biologically inspired multi-scale processing and attention mechanisms.
Why It Matters
Accurate polyp segmentation is crucial for cancer prevention during colonoscopy. GRAFNet addresses common challenges in existing methods, such as false positives and negatives, by leveraging principles of human visual processing, thus improving clinical outcomes and AI interpretability.
Key Takeaways
- GRAFNet employs a Guided Asymmetric Attention Module to enhance polyp boundary detection.
- The architecture includes a MultiScale Retinal Module for effective multi-feature analysis.
- GRAFNet demonstrates 3-8% improvements in Dice scores over leading segmentation methods.
- The model provides interpretable decision pathways, bridging AI accuracy with clinical trust.
- Extensive testing on five benchmarks confirms GRAFNet's state-of-the-art performance.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.15072 (cs) [Submitted on 15 Feb 2026] Title:GRAFNet: Multiscale Retinal Processing via Guided Cortical Attention Feedback for Enhancing Medical Image Polyp Segmentation Authors:Abdul Joseph Fofanah, Lian Wen, Alpha Alimamy Kamara, Zhongyi Zhang, David Chen, Albert Patrick Sankoh View a PDF of the paper titled GRAFNet: Multiscale Retinal Processing via Guided Cortical Attention Feedback for Enhancing Medical Image Polyp Segmentation, by Abdul Joseph Fofanah and 5 other authors View PDF Abstract:Accurate polyp segmentation in colonoscopy is essential for cancer prevention but remains challenging due to: (1) high morphological variability (from flat to protruding lesions), (2) strong visual similarity to normal structures such as folds and vessels, and (3) the need for robust multi-scale detection. Existing deep learning approaches suffer from unidirectional processing, weak multi-scale fusion, and the absence of anatomical constraints, often leading to false positives (over-segmentation of normal structures) and false negatives (missed subtle flat lesions). We propose GRAFNet, a biologically inspired architecture that emulates the hierarchical organisation of the human visual system. GRAFNet integrates three key modules: (1) a Guided Asymmetric Attention Module (GAAM) that mimics orientation-tuned cortical neurones to emphasise polyp boundaries, (2) a MultiScale Retinal Module (MSRM) that replicates retina...
