Deep learning Cancer Diagnosis
Principal Investigator
Name
Kang Li
Degrees
Ph.D
Institution
Harbin Medical University
Position Title
Professor
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-826
Initial CDAS Request Approval
Aug 6, 2021
Title
Deep learning Cancer Diagnosis
Summary
Lung cancer is by far the leading cause of cancer death among both men and women worldwide, making up almost 25% of all cancer deaths. About 80% to 85% of lung cancers are non-small cell lung cancer (NSCLC). The two most prevalent histological types of NSCLC are lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), and each is associated with very different treatment guidelines. Specifically, the available treatment options differ for LUAD and LUSC. Therefore, reliable diagnostic decision support tools are highly demanded to empower pathologists’ efficiency and accuracy to ultimately provide better patient care. Here, we develop a novel attention- and graph-guided weakly supervised deep learning model. It is unfeasible to rely on supervised learning, which requires manual annotations. Instead, we propose to use the slide-level diagnosis, which is readily available, to train an end-to-end classification model in a weakly supervised manner.
Aims
• We propose a novel end-to-end model for weakly supervised classification of lung cancer subtype.
• We empirically show that our model is robust to the existing algorithms.
• We validated the model in multi-center datasets.
Collaborators
Yan Hou, Peking University
Mantang Qiu, Peking University