Detection of Metastatic Breast Cancer from Whole-Slide Pathology Images Using an Ensemble Deep-Learning Method
Problem
Detecting lymph node metastases from whole-slide pathology images, which is time-consuming with current methods.
Why it matters
Metastasis is stated as the main cause of death among breast cancer patients, so faster detection matters.
What the AI did
Trained hybrid convolutional deep-learning models on 220,025 whole-slide images to automatically identify and quantify metastatic lymph nodes, verified on 57,458 images.
Results
98.84% accuracy for quantification; VGG16 precision 92.42% and recall 91.25%.
Impact
Demonstrated an automated deep-learning system for metastatic lymph node identification with high reported accuracy.
Record
| Venue | Archives of Breast Cancer |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Jafar Abdollahi, Niyousha Davari, Yasin Panahi, Mossa Gardaneh |
| Citations | 46 (OpenAlex, at import time) |
| References | 34 |
| Model or method | Hybrid convolutional network models; VGG16 |
| Organization | not stated in source |
| Primary source | https://doi.org/10.32768/abc.202293364-376 |
| Open access | diamond |
| Keywords | breast cancer, metastasis, deep learning, whole-slide images, VGG16, pathology |
Importance rating
major — millions of lives, an industry or a field
Abstract states metastasis is the main cause of death among breast cancer patients; automated detection could affect large patient populations.