Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 23 min ago
Peer-reviewedCC BYScientific Reports

Retinal vessel segmentation using Kolmogorov-Arnold networks and test-time adaptation

  • Haixia Bai1
  • Fuquan Wu2
  • Longyi Sun3
  • Shijing Wu1
  • Ailing Sui1
  • Zizhao Wu3
  • Zhiqing Chen1
  1. 1Eye Center of Second Affiliated Hospital, School of Medicine, Zhejiang University
  2. 2The First Affiliated Hospital of Zhejiang Chinese Medical University (Zhejiang Provincial Hospital of Chinese Medicine)
  3. 3School of Digital Media Technology, Hangzhou Dianzi University
Read full textDOICite

Abstract

Recently, TransUNet has established itself as a state-of-the-art architecture for medical image segmentation. Although attention mechanisms have yielded remarkable advances, its Multi-Layer Perceptron (MLP)-based design still suffers from limitations in capturing the complex non-linear representations essential for segmenting intricate anatomical structures. Additionally, its static inference paradigm lacks robustness when confronted with domain shifts, a common challenge in medical image data. To address these inherent limitations, we propose a dual-optimized segmentation framework, drawing inspiration from the superior non-linear approximation capability of Kolmogorov–Arnold Networks (KAN) and the adaptive advantage of Test-Time Adaptation (TTA). Specifically, we redesign the feature transformation pathway within the Transformer encoder by integrating learnable spline-based mappings derived from KAN, which significantly enhances the model’s representation capacity. Furthermore, we reformulate the inference process by incorporating a TTA strategy that enables online parameter adaptation through the minimization of prediction entropy during deployment. Extensive experiments conducted on retinal vessel segmentation benchmarks demonstrate that the proposed method consistently outperforms state-of-the-art approaches, achieving substantial improvements in both segmentation accuracy and robustness against domain variations. Our implementation is available at: https://github.com/plo404/KAN-TransUnet.

Related

  1. Evaluating the real-world accessibility of open-source deep learning models for retinal vessel segmentation in optical coherence tomography angiography
    Dayu Liu, Madeleine S. Torcasso, David Alonso-Caneiro +8
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textMedical imaging & radiologyCrossref
  2. Automated Segmentation of Retinal Pigment Epithelium Layers in High-Resolution OCT
    Giulia Ritter, Ben Isselmann, Marlene Saßmannshausen +6
    Ophthalmology sciencePeer-reviewedCC BYFull textGeneral medicineEurope PMC
  3. Topology-preserving coronary vessel segmentation in X-ray angiography using ResNet50-MASAG-UNet.
    G Sunilkumar, P Kumaresan
    Frontiers in radiologyPeer-reviewedCC BYFull textMedical imaging & radiologyEurope PMCPMC
  4. Structure-aware learnable multi-scale attention for retinal vessel analysis
    Dinoy Johny, V. Chandan, Shubham Loni +6
    Scientific ReportsPeer-reviewedCC BYFull textMedical imaging & radiologyDOAJ
  5. Artificial intelligence for ophthalmic vessel segmentation: from multimodal imaging to clinical translation
    Lingyan Wu, Jiahao Tang, Shanping Zhao
    Frontiers in medicinePeer-reviewedCC BYFull textMedical imaging & radiologyEurope PMCPMCEurope PMC

Cite

BibTeXRIS
BibTeX
@article{bai2026retinal,
  title = {{Retinal vessel segmentation using Kolmogorov-Arnold networks and test-time adaptation}},
  author = {Haixia Bai and Fuquan Wu and Longyi Sun and Shijing Wu and Ailing Sui and Zizhao Wu and Zhiqing Chen},
  journal = {Scientific Reports},
  year = {2026},
  month = jul,
  doi = {10.1038/s41598-026-61199-x},
  url = {https://doi.org/10.1038/s41598-026-61199-x},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Retinal vessel segmentation using Kolmogorov-Arnold networks and test-time adaptation
AU  - Haixia Bai
AU  - Fuquan Wu
AU  - Longyi Sun
AU  - Shijing Wu
AU  - Ailing Sui
AU  - Zizhao Wu
AU  - Zhiqing Chen
PY  - 2026
DA  - 2026/07/01
JO  - Scientific Reports
DO  - 10.1038/s41598-026-61199-x
UR  - https://doi.org/10.1038/s41598-026-61199-x
AB  - Recently, TransUNet has established itself as a state-of-the-art architecture for medical image segmentation. Although attention mechanisms have yielded remarkable advances, its Multi-Layer Perceptron (MLP)-based design still suffers from limitations in capturing the complex non-linear representations essential for segmenting intricate anatomical structures. Additionally, its static inference paradigm lacks robustness when confronted with domain shifts, a common challenge in medical image data. To address these inherent limitations, we propose a dual-optimized segmentation framework, drawing inspiration from the superior non-linear approximation capability of Kolmogorov–Arnold Networks (KAN) and the adaptive advantage of Test-Time Adaptation (TTA). Specifically, we redesign the feature transformation pathway within the Transformer encoder by integrating learnable spline-based mappings derived from KAN, which significantly enhances the model’s representation capacity. Furthermore, we reformulate the inference process by incorporating a TTA strategy that enables online parameter adaptation through the minimization of prediction entropy during deployment. Extensive experiments conducted on retinal vessel segmentation benchmarks demonstrate that the proposed method consistently outperforms state-of-the-art approaches, achieving substantial improvements in both segmentation accuracy and robustness against domain variations. Our implementation is available at: https://github.com/plo404/KAN-TransUnet.
SN  - 2045-2322
ER  -