Robust disease prognosis via diagnostic knowledge preservation: A sequential learning approach

Haresh Rengaraj Rajamohan, Yanqi Xu, Weicheng Zhu, Richard Kijowski, Kyunghyun Cho, Krzysztof J Geras, Narges Razavian, Cem M Deniz · 2026

Read Robust disease prognosis via diagnostic knowledge preservation: A sequential learning approach by Haresh Rengaraj Rajamohan, Yanqi Xu, Weicheng Zhu, Richard Kijowski, Kyunghyun Cho, Krzysztof J Geras, Narges Razavian, 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/plos-pone-0344600.txt.

Description

Haresh Rengaraj Rajamohan and co-workers teach medical image models to predict a disease's future while keeping their skill at spotting it now. Prognosis data are scarce, but diagnostic scans pile up in archives. Their 2026 study pre-trains models on large sets of knee, brain, and breast images labeled for current disease, then adapts them to forecast worsening years ahead.

Pre-training lifted forecast accuracy across all three diseases compared with standard starts. Plain fine-tuning, though, made models forget diagnosis, a safety risk in live clinics. Their step-by-step method with replay of old cases kept both skills at once, matching specialist models on prognosis while holding diagnostic accuracy. Hospitals could reuse the same trusted tool instead of juggling two.

Catalog record

CREATORHaresh Rengaraj Rajamohan, Yanqi Xu, Weicheng Zhu, Richard Kijowski, Kyunghyun Cho, Krzysztof J Geras, Narges Razavian, Cem M Deniz
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40843
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0344600

Browse the full library · About the Tower of Babel