Predicting gastric cancer outcome from resected lymph node histopathology images using deep learning
Problem
Pathologists visually inspect whole-slide lymph node images to count metastatic lymph nodes, yet patients at the same N stage still have dramatically different outcomes.
Why it matters
N-staging drives prognostic assessment and stage-based cancer therapy decisions.
What the AI did
A deep-learning framework analyzes lymph node whole-slide images to identify lymph nodes and tumor regions and derive the tumor-area-to-MLN-area ratio (T/MLN).
Results
Tumor detection performance comparable to experienced pathologists; similar performance on two independent gastric cancer validation cohorts; T/MLN shown as an interpretable independent prognostic factor.
Impact
Could assist pathologists in detecting metastatic lymph nodes and oncologists in prognostic assessment.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Xiaodong Wang, Ying Chen, Yunshu Gao, Huiqing Zhang, Zehui Guan, Zhou Dong, Yuxuan Zheng, Jiarui Jiang et al. |
| Citations | 186 (OpenAlex, at import time) |
| References | 39 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-021-21674-7 |
| Open access | gold |
| Keywords | gastric cancer, lymph nodes, deep learning, whole-slide images, prognosis |
Importance rating
significant — thousands to millions affected
N-staging drives prognosis and therapy decisions; model matches experienced pathologists on two gastric cancer cohorts and yields a prognostic factor, but abstract gives no patient-volume scale.