Federated continual learning for privacy-preserving chest radiograph classification
Anay Sinhal, Amit Sinhal, Arpana Sinhal · 2026
Read Federated continual learning for privacy-preserving chest radiograph classification by Anay Sinhal, Amit Sinhal, Arpana Sinhal (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/PMC13500843.txt.
Description
Anay Sinhal, Amit Sinhal, and Arpana Sinhal present a 2026 method for training chest X-ray classifiers across hospitals without moving patient images. Two problems meet here: privacy law blocks pooling raw scans, and models updated on new tasks tend to forget older findings. Their system, DP-FedEPC, combines federated averaging with elastic weight consolidation, which penalizes changes to weights that mattered for earlier tasks, and a small bank of class prototypes rather than stored images. Each hospital trains under differentially private stochastic gradient descent, which adds noise to protect individuals.
Training used CheXpert data, with outside checks on MIMIC-CXR. Beyond macro-AUROC accuracy, the authors report per-finding scores, forgetting scores, prototype geometry, failure cases, and a cost comparison with baselines. The design targets hospital settings where neither public substitute data nor image replay is acceptable.
Catalog record
| CREATOR | Anay Sinhal, Amit Sinhal, Arpana Sinhal |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 90932 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |