Findings from two previous trials demonstrated the efficacy of lung cancer screening (LCS) using low-dose computed tomography (LDCT) in cigarette smokers, with 20% and 24% decreases, respectively; however, innovative measures to screen and accurately predict the risk of lung cancer across a broader patient population are warranted.

In a recent publication in the Journal of Clinical Oncology, researchers from Harvard Medical School at the Massachusetts General Hospital (MGH), in collaboration with researchers at the Massachusetts Institute of Technology, theorized that a deep learning model (DLM) evaluating the entire volumetric LDCT data could be constructed to predict individual risk for lung cancer without the need for other data, including demographic or clinical data.

Employing data from LDCTs from the National Lung Screening Trial (NLST), the
researchers created a DLM named “Sybil” that evaluates, scans, and predicts the risk of lung cancer for the next 1 to 6 years. This innovative DLM needs only one LDCT, does not require clinical data or radiologist interpretations, and can run in real-time in the background on a radiology reading station.

The researchers indicated that the use of Sybil was confirmed by employing three independent data sets, including a data set of 13,309 LDCTs from a set of 6,282 LDCTS from NLST participants; 8,821 LDCTs from MGH between 2015 and 2021, and 12,280 LDCTs from adult patients who had undergone LDCTs for LCS at Chang Gung Memorial Hospital in Taiwan between 2007 and 2019. The latter dataset differed from the NLST and MGH groups because it included individuals with an assortment of smoking histories, including nonsmokers.

The results revealed that across the data sets, Sybil was capable of accurately predicting the risk of lung cancer. The researchers determined Sybil’s accuracy by utilizing the area under the curve (AUC), of which 1.0 is a perfect score.

The results also revealed that Sybil predicted cancer within 1 year with AUCs of 0.92 for the additional NLST participants, 0.86 for the MGH data set, and 0.94 for the data set from Taiwan. Moreover, Sybil predicted lung cancer within 6 years with AUCs of 0.75, 0.81, and 0.80 for the three data sets, respectively.

Lecia Sequist, the HMS Landry Family professor of medicine in the field of medical oncology at MGH, stated, “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.”

Coauthor and Jameel Clinic faculty lead, Regina Barzilay, a member of the Koch Institute for Integrative Cancer Research, stated, “Sybil can look at an image and predict the risk of a patient developing lung cancer within 6 years. I am excited about translational efforts led by the MGH team that are aiming to change outcomes for patients who would otherwise develop advanced diseases.”

The researchers stated that this is a retrospective study, and prospective studies that follow patients in the future are necessary to validate Sybil. Moreover, Dr. Sequist and colleagues will launch a prospective clinical trial to assess the capability of Sybil in the real world and understand how its use will aid radiologists.

The content contained in this article is for informational purposes only. The content is not intended to be a substitute for professional advice. Reliance on any information provided in this article is solely at your own risk.

« Click here to return to Lung Cancer Awareness.