MedSymmFlow: Bridging Generative Modeling and Classification in Medical Imaging through Symmetrical Flow Matching
Francisco Caetano, Lemar Abdi, Christiaan Viviers, Amaan Valiuddin, Fons van der Sommen · 2025
Read MedSymmFlow: Bridging Generative Modeling and Classification in Medical Imaging through Symmetrical Flow Matching by Francisco Caetano, Lemar Abdi, Christiaan Viviers, Amaan Valiuddin, Fons van der Sommen (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-2507-19098.txt.
Description
Francisco Caetano, Lemar Abdi, Christiaan Viviers, Amaan Valiuddin, and Fons van der Sommen present MedSymmFlow, a model that both classifies medical images and gauges its own doubt. Their paper from twenty twenty-five works in a high-stakes field, where a sure-sounding wrong call can harm care. The model grows from Symmetrical Flow Matching, a generative scheme. It works in a compact latent space that scales to large images.
A meaning-based mask guides the model toward features that matter for diagnosis. Unlike plain classifiers, it draws samples to sense its own uncertainty. Tests span four MedMNIST sets with mixed scan types and diseases. Accuracy and AUC matched or beat set baselines. Doubt scores proved sound: when the model could pass on hard cases, scores rose. The design unites sorting, sample-making, and doubt in one tool.
Catalog record
| CREATOR | Francisco Caetano, Lemar Abdi, Christiaan Viviers, Amaan Valiuddin, Fons van der Sommen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-19098