Washington: Lung cancer it is the leading cause of cancer death in the United States and throughout the world. Low-dose chest computed tomography (LDCT) is suggested for follow-up in people 50 to 80 years of age with a significant history of smoking or who are active smokers. Lung cancer screening with LDCT has been shown to reduce death from lung cancer by up to 24 percent.
But as lung cancer rates among non-smokers rise, new strategies are needed to accurately detect and predict lung cancer risk in a broader population. A study led by researchers at the Mass General Cancer Center, a member of Mass General Brigham, in collaboration with researchers at the Massachusetts Institute of Technology (MIT), developed and tested an artificial intelligence tool known as Sybil. Based on analyzes of LDCT scans of patients in the US and Taiwan, Sybil accurately predicted lung cancer risk for people with or without a significant history of smoking. The results are published in the Journal of Clinical Oncology.
“Lung cancer rates continue to rise among people who have never smoked or have not smoked in years, suggesting that there are many risk factors that contribute to lung cancer risk, some of which are currently unknown.” said corresponding author Lecia Sequist, MD. , MPH, director of the Center for Innovation in Cancer Early Detection and lung cancer medical oncologist at the Mass General Cancer Center. “Instead of assessing individual environmental or genetic risk factors, we have developed a tool that can use imaging to look at collective biology and make predictions about cancer risk.”
The US Preventive Service Task Force recommends annual LDCTs for people age 50 and older with a 20-pack-year history, who are current smokers or have quit within the past 15 years. But less than 10 percent of eligible patients are screened annually. To help improve the efficiency of lung cancer screening and provide individualized evaluations, Sequist and his colleagues at the Mass General Cancer Center partnered with researchers at the MIT Jameel Clinic. Using data from the National Lung Screening Trial (NLST), the team developed Sybil, a deep-learning model that analyzes scans and predicts lung cancer risk for the next one to six years.
“Sybil requires only LDCT and is not dependent on clinical data or radiologist notes,” said co-author Florian Fintelmann, MD, Department of Radiology, Division of Thoracic Imaging and Intervention, Massachusetts General Hospital. “It was designed to run in real time in the background of a standard radiology reading station enabling clinical decision support at the point of care.”
The team validated Sybil using three independent data sets: a set of scans of more than 6,000 NLST participants that Sybil had not seen before; 8,821 LDCT from Massachusetts General Hospital (MGH); and 12,280 LDCT from Chang Gung Memorial Hospital in Taiwan. The latest set of scans included people with a variety of smoking histories, including those who never smoked.
Sybil was able to accurately predict the risk of lung cancer in these sets. The researchers determined Sybil’s precision using Area Under the Curve (AUC), a measure of how well a test can distinguish between diseased and normal samples, with 1.0 being a perfect score. Sybil predicted cancer within one year with AUC of 0.92 for the additional NLST participants, 0.86 for the MGH data set, and 0.94 for the Taiwan data set. The program predicted lung cancer within six years with AUCs of 0.75, 0.81, and 0.80, respectively, for the three data sets.
“Sybil can look at an image and predict a patient’s risk of developing lung cancer within six years,” said co-author and Jameel Clinic faculty leader Regina Barzilay, PhD, a member of the Koch Institute for Research. Cancer Integrative. “I am excited about the translational efforts led by the MGH team that aim to change outcomes for patients who would otherwise develop advanced disease.”
The researchers note that this is a retrospective study and prospective studies that follow patients in the future are needed to validate Sybil. In addition, the American participants in the study were overwhelmingly white (92 percent), and future studies will be needed to determine if Sybil can accurately predict lung cancer among diverse populations.
Sequist and his colleagues will open a prospective clinical trial to test Sybil in the real world 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 risk patterns from the LDCT that were not visible to the naked eye,” Sequist said. “We are excited to continue testing this program to see if it can add information that will help radiologists with diagnosis and put us on the path to personalize screening tests for patients.”
