Study
NLST
(Learn more about this study)
Project ID
NLST-772
Initial CDAS Request Approval
Mar 22, 2021
Title
Predicting outcome of lung cancer patients using deep learning
Summary
In this project, we intend to utilise deep learning to develop a system for prediction of patient outcome directly from scanned conventional haematoxylin and eosin (H&E) stained sections. We have successfully developed such a biomarker for patients resected for early-stage colorectal cancer and published our findings in The Lancet in 2020 (
https://doi.org/10.1016/S0140-6736(19)32998-8 ). The marker was validated in an external cohort according to a pre-specified protocol and performed better than most other markers in terms of hazard ratios in stage-specific multivariable analyses. We now seek datasets to repeat the endeavour in lung cancer and therefore wish to gain access to the NLST Pathology Images with all associated data, in particular clinical, pathological and follow-up data. The ultimate aim of the project is to provide the clinic with an objective, automatic and low-cost tool that can enable clinicians and patients to make joint and more informed decisions on adjuvant treatment options after resection of lung cancer.
Aims
- Develop and validate a system for prediction of patient outcome directly from scanned conventional H&E stained sections of lung cancer resection specimens.
- If successful, the ultimate goal is to improve treatment of lung cancer patients through the use of an automatic and objective prognostic marker that supplements established markers.
Collaborators
Ole-Johan Skrede, Institute for Cancer Genetics and Informatics, Oslo University Hospital
Tarjei Sveinsgjerd Hveem, Institute for Cancer Genetics and Informatics, Oslo University Hospital
Knut Liestøl, Institute for Cancer Genetics and Informatics, Oslo University Hospital
Manohar Pradhan, Institute for Cancer Genetics and Informatics, Oslo University Hospital
Odd Terje Brustugun, Drammen Hospital, Vestre Viken Hospital Trust
