학술논문
The impact of site-specific digital histology signatures on deep learning model accuracy and bias
Document Type
article
Author
Source
Nature Communications, Vol 12, Iss 1, Pp 1-13 (2021)
Subject
Language
English
ISSN
2041-1723
Abstract
Deep learning models have been trained on The Cancer Genome Atlas to predict numerous features directly from histology, including survival, gene expression patterns, and driver mutations. Here, the authors demonstrate that site-specific histologic signatures can lead to biased estimates of accuracy for such models, and propose a method to minimize such bias.