TCSAFormer: Efficient Vision Transformer with Token Compression and Sparse Attention for Medical Image Segmentation

December 30, 2025 · View on GitHub

Zunhui Xia, Hongxing Li, Libin Lan

Performance comparison of different medical image segmentation models：

TCSAFormer Architecture:

Compression Attention:

Usage

Environment

  • python == 3.10
  • torch == 2.0.0+cu118
  • torchaudio == 2.0.1+cu118

Contents

  1. 1Performance comparison of different medical image segmentation models：
  2. 2TCSAFormer Architecture:
  3. 3Compression Attention:
  4. 4Usage
  5. 4.1Environment