Building a General SimCLR Self-Supervised Foundation Model Across Neurological Diseases to Advance 3D Brain MRI Diagnoses
Emily Kaczmarek, Justin Szeto, Brennan Nichyporuk, Tal Arbel, Mila - Quebec Artificial Intelligence Institute, Canada, [email protected], [email protected] · 2025
Read Building a General SimCLR Self-Supervised Foundation Model Across Neurological Diseases to Advance 3D Brain MRI Diagnoses by Emily Kaczmarek, Justin Szeto, Brennan Nichyporuk, Tal Arbel, Mila - Quebec Artificial Intelligence Institute, Canada, [email protected], [email protected] (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-2509-10620.txt.
Description
Emily Kaczmarek and three co-authors build a foundation model for brain MRI in this 2025 medical AI paper. Deep learning reads brain scans well but usually for one narrow task at a time, starved of labeled data. Their model instead learns from unlabeled scans first, using SimCLR, a self-supervised method that teaches a network by matching different views of the same image.
Pretraining covered nearly nineteen thousand patients and forty-five thousand scans from eleven public datasets spanning many neurological diseases, at high resolution in full 3D. Against masked autoencoders and supervised baselines, the fine-tuned model won on all four test tasks, in familiar and shifted populations alike. For Alzheimer's prediction it stayed ahead even when fine-tuned on only a fifth of the labels. The authors release the model and code publicly.
Catalog record
| CREATOR | Emily Kaczmarek, Justin Szeto, Brennan Nichyporuk, Tal Arbel, Mila - Quebec Artificial Intelligence Institute, Canada, [email protected], [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-10620