{"ID":2894065,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.12427","arxiv_id":"2507.12427","title":"Unit-Based Histopathology Tissue Segmentation via Multi-Level Feature Representation","abstract":"We propose UTS, a unit-based tissue segmentation framework for histopathology that classifies each fixed-size 32 * 32 tile, rather than each pixel, as the segmentation unit. This approach reduces annotation effort and improves computational efficiency without compromising accuracy. To implement this approach, we introduce a Multi-Level Vision Transformer (L-ViT), which benefits the multi-level feature representation to capture both fine-grained morphology and global tissue context. Trained to segment breast tissue into three categories (infiltrating tumor, non-neoplastic stroma, and fat), UTS supports clinically relevant tasks such as tumor-stroma quantification and surgical margin assessment. Evaluated on 386,371 tiles from 459 H\u0026E-stained regions, it outperforms U-Net variants and transformer-based baselines. Code and Dataset will be available at GitHub.","short_abstract":"We propose UTS, a unit-based tissue segmentation framework for histopathology that classifies each fixed-size 32 * 32 tile, rather than each pixel, as the segmentation unit. This approach reduces annotation effort and improves computational efficiency without compromising accuracy. To implement this approach, we introd...","url_abs":"https://arxiv.org/abs/2507.12427","url_pdf":"https://arxiv.org/pdf/2507.12427v1","authors":"[\"Ashkan Shakarami\",\"Azade Farshad\",\"Yousef Yeganeh\",\"Lorenzo Nicole\",\"Peter Schüffler\",\"Stefano Ghidoni\",\"Nassir Navab\"]","published":"2025-07-16T17:15:18Z","proceeding":"eess.IV","tasks":"[\"eess.IV\",\"cs.AI\",\"cs.CV\",\"cs.LG\"]","methods":"[\"Vision Transformer\",\"Transformer\"]","has_code":false}