Skip to main navigation Skip to search Skip to main content

KAN-OSD: Dino-Based Encoder and KAN-Based Decoders with Dual Contrastive Learning for One-Shot Anatomical Landmark Detection

  • Yinbin Tian
  • , Ziyang Wang
  • , Shibiao Xu
  • , Li Guo*
  • *Corresponding author for this work
  • Beijing University of Posts and Telecommunications

Research output: Chapter in Book/Published conference output › Conference publication

5 Downloads (Pure)

Abstract

One-shot anatomical landmark detection seeks accurate localization from a single labeled template. To address this data scarcity issue, an effective strategy is to leverage the strong generalization of a ViT-based foundation-model encoder, particularly its out-of-domain knowledge. We propose KANOSD, which freezes the encoder and trains only a lightweight prediction head with distance-aware similarity and contrastive objectives. Conventional convolutional decoders are replaced by KolmogorovArnold Networks (KAN) to model a continuous, high-resolution similarity field conditioned jointly on foundation-model features and image coordinates. To further enhance discriminability under limited supervision, we introduce metric-aware contrastive learning, which uses multiscale and multi-view augmentations of the template image as network inputs. Contrastive learning across global and local features strengthens context consistency for each landmark, while diverse augmentation views help reduce overfitting and improve generalization. KAN-OSD maintains strong label and computational efficiency without requiring any unlabeled images, while achieving lower mean radial error and higher successful detection rates on Head and Hand X-ray benchmarks.
Original languageEnglish
Title of host publication2026 IEEE 23rd International Symposium on Biomedical Imaging (ISBI)
PublisherIEEE
Number of pages4
ISBN (Electronic)9798331577636
DOIs
Publication statusPublished - 20 May 2026

Publication series

NameSymposium Proceedings: The International Symposium on Biomedical Imaging
PublisherIEEE
ISSN (Electronic)1945-8452

Bibliographical note

This is an accepted manuscript of an article published in: 2026 IEEE 23rd International Symposium on Biomedical Imaging (ISBI) . For the purposes of open access the author/s has/ve applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript (AAM) version arising from this submission.

Funding

This work was supported in part by Key Research and Development Program of Hainan Province under Grant ZDYF2025 (LALH) 002, and in part by the Beijing Natural Science Foundation under Grant L232039.

FundersFunder number
Key Research and Development Program of Hainan ProvinceZDYF2025 (LALH) 002
Natural Science Foundation of Beijing MunicipalityL232039

    Keywords

    • Contrastive Learning
    • Few-shot Learning
    • Kolmogorov-Arnold Networks
    • Landmark Detection

    Fingerprint

    Dive into the research topics of 'KAN-OSD: Dino-Based Encoder and KAN-Based Decoders with Dual Contrastive Learning for One-Shot Anatomical Landmark Detection'. Together they form a unique fingerprint.

    Cite this