To develop the artificial intelligence-based approach, the research team used chest radiograph images and patient demographic data including age, sex and current smoking status that is available in EHRs. The model makes predictions based on chest X-ray images from more than 41,000 people that participated in a multicenter trial of lung cancer screening with chest X-rays, called the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial.
Researchers validated the final model in 5,615 additional trial participants and 5,493 individuals involved from a second trial, the National Lung Screening Trial. The deep learning model outperformed the standard Medicare lung cancer screening criteria, which is the current clinical standard, and missed 30.7 percent fewer lung cancers while screening the same number of individuals.
More articles on artificial intelligence:
Penn State develops algorithm to alert users of illegitimate medications online
Penn Medicine AI assists global consortium’s COVID-19 data model: 4 notes
Defense Health Agency launches AI project for cancer detection
