Mix, Align, Distil: Reliable Cross-Domain Atypical Mitosis Classification
Kaustubh Atey, Sameer Anand Jha, Gouranga Bala, Amit Sethi · 2025
Read Mix, Align, Distil: Reliable Cross-Domain Atypical Mitosis Classification by Kaustubh Atey, Sameer Anand Jha, Gouranga Bala, Amit Sethi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-20745.txt.
Description
Spotting abnormal cell division in tissue slides is a marker task in cancer care. A 2025 paper by Kaustubh Atey and co-authors offers a training recipe for it. Odd mitotic figures are hard to grade the same way across labs, since scanners, stains, and sample sources all differ.
Their recipe has three parts. Style changes added during training widen the range of looks. A loss aligns features across sites using rough labels. A teacher model steadies the student's answers. On the early board for the MIDOG 2025 task, the entry reached a balanced accuracy of about 0.88, with strong scores on both classes. The method adds almost no cost when the model is used.
Catalog record
| CREATOR | Kaustubh Atey, Sameer Anand Jha, Gouranga Bala, Amit Sethi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 14 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-20745