A novel sub-differentiable hausdorff loss combined with BCE for MRI brain tumor segmentation using UNet variants
Gurugubelli Visalakshi, Laavanya Mohan · 2025
Read A novel sub-differentiable hausdorff loss combined with BCE for MRI brain tumor segmentation using UNet variants by Gurugubelli Visalakshi, Laavanya Mohan (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/pmc-12749605.txt.
Description
Gurugubelli Visalakshi and Laavanya Mohan sharpen the outlines that software draws around brain tumors in a 2025 paper. Tumor pixels are a tiny share of a brain scan, so models tend to miss small or ragged tumors and blur their edges.
Standard scoring rules punish those edge errors weakly, and a boundary focused rule in wide use is unstable during training. Their new loss rule keeps the boundary focus but is smoothed so training stays steady. Mixed with a region based rule, it gives both crisp edges and complete coverage. Five network designs, all from one popular family, were trained with the new losses and scored on overlap and accuracy measures.
Catalog record
| CREATOR | Gurugubelli Visalakshi, Laavanya Mohan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 144 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |