Self-supervised attention-based deep learning for pan-cancer mutation prediction from histopathology
Problem
It was unclear whether deep learning predictions of genetic alterations from tumor histopathology slides generalize to external datasets across cancer types.
Why it matters
Predicting mutations from routine pathology slides could inform cancer care without genetic testing.
What the AI did
A pipeline integrating self-supervised feature extraction with attention-based multiple instance learning predicted genetic alterations from histology on two large multi-tumor datasets.
Results
Robust predictability and generalizability claimed on two large datasets; no numeric metrics in abstract.
Impact
The study demonstrates externally generalizable AI prediction of cancer mutations from histology, per the authors.
Record
| Venue | npj Precision Oncology |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Oliver Lester Saldanha, Chiara Maria Lavinia Loeffler, Jan Niehues, Marko van Treeck, Tobias Paul Seraphin, Katherine Jane Hewitt, Didem Çifçi, Gregory Patrick Veldhuizen et al. |
| Citations | 87 (OpenAlex, at import time) |
| References | 43 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41698-023-00365-0 |
| Open access | gold |
| Keywords | histopathology, mutation prediction, self-supervised learning, attention, pan-cancer |
Importance rating
major — millions of lives, an industry or a field
Cancer mutation prediction from routine slides could inform care across multiple tumor types; abstract cites two large multi-tumor datasets but no scale numbers.