Deep learning-based breast cancer grading and survival analysis on whole-slide histopathology images
Problem
Pathologist grading of breast cancer shows significant inter-observer variation, and prior computer methods only work on selected tissue areas or need labor-intensive annotations.
Why it matters
Tumor grade is strongly associated with patient survival, so grading consistency affects clinical decisions.
What the AI did
Trained a deep learning model to grade breast cancer directly from whole-slide histopathology images, including nuclear grade, tubule formation and mitotic rate, evaluated against expert pathologist annotations.
Results
Trained on whole-slide images from 706 patients; evaluated with Cohen's kappa on an independent test set of 686 patients (kappa value not stated in source)
Impact
Peer-reviewed evaluation on an independent 686-patient test set suggests the model can reduce reliance on manual grading, though clinical deployment is not stated.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Suzanne C. Wetstein, Vincent M. T. de Jong, Nikolas Stathonikos, Mark Opdam, Gwen M. H. E. Dackus, Josien P. W. Pluim, P. J. van Diest, Mitko Veta |
| Citations | 103 (OpenAlex, at import time) |
| References | 65 |
| Model or method | deep learning (specific architecture not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-022-19112-9 |
| Open access | gold |
| Keywords | deep learning, breast cancer, whole-slide images, tumor grading, Cohen's kappa, survival analysis |
Importance rating
significant — thousands to millions affected
Tumor grade strongly associated with survival; inter-observer variation affects clinical decisions; evaluated on 706 training and 686 independent test patients.