Artificial DNA and Intelligence
Credit: iStock/agsandrew

Researchers at Massachusetts General Hospital have developed an artificial intelligence tool called “Sybil” that can accurately predict the risk of lung cancer in individuals with or without a significant smoking history for up to six years.

The American Cancer Society estimates over 100,000 new deaths from lung cancer in the U.S. in 2023. Current standard of screening for the disease is low-dose chest computed tomography (LDCT) which is recommended for people between 50 and 80 years of age with a significant history of smoking. However, as lung cancer cases are starting to rise among non-smokers, new screening strategies are needed.

Reporting in the Journal of Clinical Oncology, researchers at Massachusetts General Hospital have developed and tested an artificial intelligence tool called “Sybil” using data from the National Lung Screening Trial, a randomized multicenter study comparing LDCT with chest radiography in the screening of individuals for early detection of lung cancer.

The scientists hypothesized that LDCT images contained information that was predictive of future lung cancer risk beyond currently identifiable features such as lung nodules. Sybil uses LDCT scan images to correctly localize the location of future cancers and determine the likelihood that an LDCT scan can be considered high-risk by recognizing pathogenic structures that are not visible to the human eye.

“Sybil requires only one LDCT and does not depend on clinical data or radiologist annotations. It was designed to run in real-time in the background of a standard radiology reading station which enables point-of care clinical decision support,” said Florian Fintelmann, MD, Department of Radiology at Massachusetts General Hospital and co-author of the study.

The researchers validated Sybil using several independent LDCT data sets including individuals with a range of smoking history as well as those without. According to the team, the artificial intelligence model was accurately able to predict the risk of lung cancer across all the data sets.

Using a test that can distinguish between disease and normal samples called “Area Under the Curve” with a maximum score of 1.0, Sybil correctly predicted lung cancer within one year with a score of 0.92 and a score of 0.75 for lung cancer risk within six years. The researchers note that the study is retrospective and that future studies with newer and more diverse data will be needed to further validate the model.

“In our study, Sybil was able to detect patterns of risk from the LDCT that were not visible to the human eye. We’re excited to further test this program to see if it can add information that helps radiologists with diagnostics and sets us on a path to personalize screening for patients,” concluded co-author and  director of the Center for Innovation in Early Cancer Detection Lecia Sequist.

Also of Interest