Researchers have developed and tested an artificial intelligence (AI) tool known as Sybil, which may accurately predict the risk of lung cancer for individuals with or without a significant smoking history using data from low-dose chest computed tomography (CT) scans, according to a new study published by Mikhael et al in the Journal of Clinical Oncology.
Lung cancer is the leading cause of cancer mortality around the world. For individuals aged 50 to 80 years who are current or former smokers, experts recommend screening with low-dose chest CT scans. Lung cancer screening with low-dose chest CT scans has been shown to reduce lung cancer mortality by up to 24%, but as rates of lung cancer climb among nonsmokers, new strategies may be needed to screen and accurately predict lung cancer risk across a wider population.
“Lung cancer rates continue to rise among [individuals] who have never smoked or who haven’t smoked in years, suggesting that there are many … factors contributing to lung cancer risk, some of which are currently unknown,” stressed corresponding study author Lecia Sequist, MD, MPH, the Landry Family Professor of Medicine at Harvard Medical School and Director of the Center for Innovation in Early Cancer Detection at the Massachusetts General Hospital Cancer Center. “Instead of assessing individual environmental or genetic risk factors, we’ve developed a tool that can use images to look at collective biology and make predictions about cancer risk.”
The U.S. Preventive Service Task Force currently recommends annual low-dose chest CT scans for individuals older than 50 with a smoking history of 20 pack-years, who either currently smoke or have quit smoking within the past 15 years. However, less than 10% of eligible patients are screened annually.
Sybil Development and Testing
To help improve the efficiency of lung cancer screening and provide individualized assessments, Dr. Sequist and her colleagues used data from the National Lung Screening Trial to develop Sybil—a deep-learning model capable of analyzing scans and predicting lung cancer risk for up to 6 years.
“Sybil requires only one [low-dose chest] CT [scan] and does not depend on clinical data or radiologist annotations,” explained study coauthor Florian Fintelmann, MD, Associate Professor of Radiology at Harvard Medical School as well as a physician-scientist in the Department of Radiology and Head of the Division of Thoracic Imaging Percutaneous Thermal Ablation at Massachusetts General Hospital. “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.”
In a retrospective study, the researchers validated Sybil using three independent data sets—scans from over 6,000 National Lung Screening Trial participants who Sybil had not previously seen, low-dose chest CT scans from 8,821 participants at Massachusetts General Hospital, and low-dose chest CT scans from 12,280 participants at Chang Gung Memorial Hospital. The latter set of scans included individuals with a range of smoking history, including never-smokers.
The researchers discovered that Sybil was able to accurately predict the risk of developing lung cancer across these sets. The researchers then determined Sybil’s predictive accuracy using area under the curve, with 1.0 as a perfect score. Sybil predicted lung cancer within 1 year with scores of 0.92 for the additional National Lung Screening Trial participants, 0.86 for the Massachusetts General Hospital participants, and 0.94 for the Chang Gung Memorial Hospital participants. Further, the program predicted lung cancer within 6 years with scores of 0.75, 0.81, and 0.80, respectively.
The researchers emphasized that future prospective studies are still needed to validate the efficacy of Sybil as well as its ability to accurately predict lung cancer among diverse populations.
Dr. Sequist and her colleagues revealed that they plan to begin a prospective clinical trial to test Sybil in a real-world setting and understand how it complements the work of radiologists. The code has also been made publicly available.
“In our study, Sybil was able to detect patterns of risk from [low-dose chest] CT [scans] that were not visible to the human eye,” concluded Dr. Sequist. “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.”
Disclosure: The research in this study was supported by the Bridge Project partnership between the Koch Institute for Integrative Cancer Research at MIT and the Dana-Farber/Harvard Cancer Center, the MIT Jameel-Clinic, Quanta Computing, Stand Up To Cancer, the Massachusetts General Hospital Center for Innovation in Early Cancer Detection, the Bralower and Landry Families, and Upstage Lung Cancer. Funding to support this research was also provided by the Eric and Wendy Schmidt Center at the Broad Institute of MIT and Harvard. For full disclosures of the study authors, visit ascopubs.org.The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.