Scholarlink

The Diabetic Foot Ulcer Challenge is a global computer vision competition organised to raise awareness of the complications caused by diabetes and motivate researchers to develop solutions that might help in the treatment. In 2024, the challenge focused on self-supervised learning methods and synthetic data, as the process of collecting and labelling medical data is a tedious one. In this research, we present a novel, two-stage method for self-supervised DFU segmentation. It is based on a Vision Transformer attention from classification pre-task, followed by U-Net weakly supervised segmentation refinement. Our solution took 3rd place on the challenge leaderboard, with a 0.4796 Dice Score on the test set.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Diabetic Foot Ulcer Unsupervised Segmentation with Vision Transformers Attention

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Andrzej Brodzicki,
  • Dariusz Kucharski,
  • Bartłomiej Moniak,
  • Aleksander Kostuch,
  • Filip Noworolnik,
  • W. Anna,
  • Joanna Jaworek-Korjakowska

摘要

The Diabetic Foot Ulcer Challenge is a global computer vision competition organised to raise awareness of the complications caused by diabetes and motivate researchers to develop solutions that might help in the treatment. In 2024, the challenge focused on self-supervised learning methods and synthetic data, as the process of collecting and labelling medical data is a tedious one. In this research, we present a novel, two-stage method for self-supervised DFU segmentation. It is based on a Vision Transformer attention from classification pre-task, followed by U-Net weakly supervised segmentation refinement. Our solution took 3rd place on the challenge leaderboard, with a 0.4796 Dice Score on the test set.