Accurate diagnosis of lymphoma on whole-slide histopathology images using deep learning
Problem
Histopathological diagnosis of lymphomas, especially follicular lymphoma, is challenging, requires expertise or centralized review, and depends on tissue-section technical quality.
Why it matters
Accurate lymphoma diagnosis determines patient treatment, and expert pathology review is not universally available.
What the AI did
Bayesian neural networks trained, validated, and tested on whole-slide H&E images diagnose follicular lymphoma versus follicular hyperplasia, with certainty estimation that also flags unfamiliar external cases.
Results
Area under the curve reaching 0.99; uncertainty estimation detected unfamiliar data such as other small B cell lymphomas and external-centre cases
Impact
Demonstrated accurate FL diagnosis with uncertainty-based detection of out-of-distribution cases in a peer-reviewed study.
Record
| Venue | npj Digital Medicine |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Charlotte Syrykh, Arnaud Abreu, Nadia Amara, Aurore Siegfried, V. Maisongrosse, François-Xavier Frénois, Laurent Martin, Cédric Rossi et al. |
| Citations | 115 (OpenAlex, at import time) |
| References | 28 |
| Model or method | Bayesian neural networks (BNN) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41746-020-0272-0 |
| Open access | gold |
| Keywords | lymphoma, whole-slide images, Bayesian neural networks, histopathology, uncertainty estimation |
Importance rating
significant — thousands to millions affected
Accurate lymphoma diagnosis guides treatment and expert pathology review is not universally available, but the abstract gives no scale numbers; benefit is significant but narrower than top-tier.