MA-KANet: Enhancing Retinal Vessel Segmentation through Multi-scale Feature Fusion and Attention Mechanisms

Guojie Li; Fan Zhang; Muhammad Ateeq; Anh Nguyen; Anwar P.P. Abdul Majeed

doi:10.1016/j.icte.2026.02.006

MA-KANet: Enhancing Retinal Vessel Segmentation through Multi-scale Feature Fusion and Attention Mechanisms

Guojie Li
, Fan Zhang
, Muhammad Ateeq
, Anh Nguyen
, Anwar P.P. Abdul Majeed^*

^*Corresponding author for this work

University of Liverpool

Research output: Contribution to journal › Article › peer-review

2 Citations (Scopus)

Abstract

We propose Multi-Scale Attention Kolmogorov–Arnold Network (MA-KANet), a novel retinal vessel segmentation framework addressing fine-scale structures and low contrast challenges. Our method integrates Multi-scale Dynamic Fusion (MDF) with 3D convolutional interactions to prevent information loss during downsampling, and Scale-Context Attention Fusion (SCAF) for dynamic feature recalibration in cluttered regions. Kolmogorov–Arnold Network units in decoder stages capture nonlinear dependencies beyond traditional convolutions. MA-KANet achieves state-of-the-art results: 98.80% AUC on DRIVE, 99.05% AUC on CHASEDB1, and 99.17% AUC on FIVES dataset, demonstrating superior performance and generalization across diverse vessel patterns, establishing new benchmarks for retinal vessel segmentation.

Original language	English
Journal	ICT Express
DOIs	https://doi.org/10.1016/j.icte.2026.02.006
Publication status	Published - 17 Feb 2026

Access to Document

10.1016/j.icte.2026.02.006Licence: CC BY-NC-ND

Advancing Ophthalmological Diagnostics Integrating Deep Learning Architectures with Fundus Imagery for Comprehensive Ocular Pathology Detection
Zhang, F. (Supervisor), Ateeq, M. (Co-supervisor), Nguyen, A. (Co-supervisor) & P.P. Abdul Majeed, A. (Co-supervisor)
2024 → 2028
Activity: Supervision › PhD Supervision

Cite this

@article{4252c69625fa4c8dbee5401fc0a8d1ee,

title = "MA-KANet: Enhancing Retinal Vessel Segmentation through Multi-scale Feature Fusion and Attention Mechanisms",

abstract = "We propose Multi-Scale Attention Kolmogorov{\textendash}Arnold Network (MA-KANet), a novel retinal vessel segmentation framework addressing fine-scale structures and low contrast challenges. Our method integrates Multi-scale Dynamic Fusion (MDF) with 3D convolutional interactions to prevent information loss during downsampling, and Scale-Context Attention Fusion (SCAF) for dynamic feature recalibration in cluttered regions. Kolmogorov{\textendash}Arnold Network units in decoder stages capture nonlinear dependencies beyond traditional convolutions. MA-KANet achieves state-of-the-art results: 98.80\% AUC on DRIVE, 99.05\% AUC on CHASEDB1, and 99.17\% AUC on FIVES dataset, demonstrating superior performance and generalization across diverse vessel patterns, establishing new benchmarks for retinal vessel segmentation.",

author = "Guojie Li and Fan Zhang and Muhammad Ateeq and Anh Nguyen and \{P.P. Abdul Majeed\}, Anwar",

year = "2026",

month = feb,

day = "17",

doi = "10.1016/j.icte.2026.02.006",

language = "English",

journal = "ICT Express",

issn = "2405-9595",

}

TY - JOUR

T1 - MA-KANet: Enhancing Retinal Vessel Segmentation through Multi-scale Feature Fusion and Attention Mechanisms

AU - Li, Guojie

AU - Zhang, Fan

AU - Ateeq, Muhammad

AU - Nguyen, Anh

AU - P.P. Abdul Majeed, Anwar

PY - 2026/2/17

Y1 - 2026/2/17

N2 - We propose Multi-Scale Attention Kolmogorov–Arnold Network (MA-KANet), a novel retinal vessel segmentation framework addressing fine-scale structures and low contrast challenges. Our method integrates Multi-scale Dynamic Fusion (MDF) with 3D convolutional interactions to prevent information loss during downsampling, and Scale-Context Attention Fusion (SCAF) for dynamic feature recalibration in cluttered regions. Kolmogorov–Arnold Network units in decoder stages capture nonlinear dependencies beyond traditional convolutions. MA-KANet achieves state-of-the-art results: 98.80% AUC on DRIVE, 99.05% AUC on CHASEDB1, and 99.17% AUC on FIVES dataset, demonstrating superior performance and generalization across diverse vessel patterns, establishing new benchmarks for retinal vessel segmentation.

AB - We propose Multi-Scale Attention Kolmogorov–Arnold Network (MA-KANet), a novel retinal vessel segmentation framework addressing fine-scale structures and low contrast challenges. Our method integrates Multi-scale Dynamic Fusion (MDF) with 3D convolutional interactions to prevent information loss during downsampling, and Scale-Context Attention Fusion (SCAF) for dynamic feature recalibration in cluttered regions. Kolmogorov–Arnold Network units in decoder stages capture nonlinear dependencies beyond traditional convolutions. MA-KANet achieves state-of-the-art results: 98.80% AUC on DRIVE, 99.05% AUC on CHASEDB1, and 99.17% AUC on FIVES dataset, demonstrating superior performance and generalization across diverse vessel patterns, establishing new benchmarks for retinal vessel segmentation.

U2 - 10.1016/j.icte.2026.02.006

DO - 10.1016/j.icte.2026.02.006

M3 - Article

SN - 2405-9595

JO - ICT Express

JF - ICT Express

ER -

MA-KANet: Enhancing Retinal Vessel Segmentation through Multi-scale Feature Fusion and Attention Mechanisms

Abstract

Access to Document

Fingerprint

Activities

Advancing Ophthalmological Diagnostics Integrating Deep Learning Architectures with Fundus Imagery for Comprehensive Ocular Pathology Detection

Cite this