A deep learning model for the classification of indeterminate lung carcinoma in biopsy whole slide images
Problem
A subset of small transbronchial lung biopsy cases cannot be confidently classified by pathologists from H&E slides alone, requiring extra immunohistochemistry or surgery.
Why it matters
Distinguishing adenocarcinoma, squamous cell carcinoma, and small-cell lung cancer determines optimum cancer treatment.
What the AI did
A deep learning model was trained on 579 whole slide images to classify lung biopsy slides into adenocarcinoma, squamous cell carcinoma, small-cell lung cancer, and non-neoplastic.
Results
AUC 0.99 on an independent test set of 83 challenging indeterminate cases; evaluated on four further independent test sets.
Impact
Peer-reviewed results suggest AI can aid pathologists on diagnostically challenging lung biopsy cases.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Fahdi Kanavati, Gouji Toyokawa, Seiya Momosaki, Hiroaki Takeoka, Masaki Okamoto, Koji Yamazaki, Sadanori Takeo, Osamu Iizuka et al. |
| Citations | 57 (OpenAlex, at import time) |
| References | 64 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-021-87644-7 |
| Open access | gold |
| Keywords | lung cancer classification, whole slide imaging, deep learning, H&E pathology, biopsy, diagnostic aid |
Importance rating
real but narrower advance
Abstract states histological differentiation is crucial for optimum lung cancer treatment; model aids on a challenging subset of biopsy cases, but no scale or clinical use stated.