A deep learning model called Sybil can predict a person’s risk of lung cancer using a single low-dose CT scan, according to data published in the Journal of Clinical Oncology.
The study showed that Sybil requires 1 low-dose CT scan with no further clinical data or radiologist annotations, and it can run in real time in the background of a radiology reading station.
The researchers validated Sybil’s performance in 3 data sets: 6282 CT scans from National Lung Screening Trial (NLST) participants, 8821 CT scans from Massachusetts General Hospital (MGH), and 12,280 CT scans from Chang Gung Memorial Hospital (CGMH).
Area under the receiver-operator curves (AUCs) for lung cancer prediction at 1 year were 0.92 for the NLST cohort, 0.86 for the MGH cohort, and 0.94 for the CGMH cohort.
Two-year AUCs were 0.86 for the NLST cohort, 0.82 for the MGH cohort, and 0.87 for the CGMH cohort. Looking further out, concordance indices over 6 years were 0.75 for the NLST cohort, 0.81 for the MGH cohort, and 0.80 for the CGMH cohort.
The researchers also performed an exploratory analysis in which Sybil’s performance was hampered by removing visible nodules. This analysis yielded a 2-year AUC of 0.81 and a 6-year AUC of 0.69 in the NLST cohort.
“On the basis of our initial results, one potential clinical application is to use Sybil to decrease follow-up scans or biopsies among patients with nodules that are low risk,” the researchers wrote.
“The preliminary results of this study suggest the program can provide additional information about the future lung cancer risk in patients undergoing CT lung cancer screening with minimal disruption in the normal clinical workflow,” JCO Associate Editor Thomas E. Stinchcombe, MD, wrote in a comment on the results. “Further evaluation in a prospective study to assess the performance and clinical benefit is warranted.”
Disclosures: This research was partly supported by Quanta Computing. Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of disclosures.
Mikhael PG, Wohlwend J, Yala A, et al. Sybil: A validated deep learning model to predict future lung cancer risk from a single low-dose chest computed tomography. J Clin Oncol. Published online January 12, 2023. doi:10.1200/JCO.22.01345