Self-Supervised Learning for Knee Osteoarthritis: Diagnostic Limitations and Prognostic Value of Hospital Data
Haresh Rengaraj Rajamohan, Yuxuan Chen, Kyunghyun Cho, Cem M. Deniz · 2026
Read Self-Supervised Learning for Knee Osteoarthritis: Diagnostic Limitations and Prognostic Value of Hospital Data by Haresh Rengaraj Rajamohan, Yuxuan Chen, Kyunghyun Cho, Cem M. Deniz (2026) 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-2603-24903.txt.
Description
Rajamohan and co-authors assess self-supervised learning for knee osteoarthritis modelling. They compare image-only self-supervision on X-rays from three cohorts with multimodal image-text pretraining on hospital images paired with notes. For diagnostic grading, the results are mixed. Image-only pretraining helps under linear probing but not under full fine-tuning, and multimodal pretraining does not improve grading. The hospital corpus covers only confirmed cases, which limits grading signal. For prognostic modelling, though, the multimodal start improves performance, the authors report.
Catalog record
| CREATOR | Haresh Rengaraj Rajamohan, Yuxuan Chen, Kyunghyun Cho, Cem M. Deniz |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53333 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.24903