A new AI solution undergoing clinical trials is capable of predicting a patient's risk for lung cancer as much as six years earlier, using indications and patterns from CT images long before human radiologists would be able to identify them.

Developed by scientists at Mass General Cancer Center and the Massachusetts Institute of Technology, Sybil is currently in clinical trials, and in one study was able to correctly predict patients' risk for developing the disease within one year 86% to 94% of the time, reported NBC News.

While the Centers for Disease Control and Prevention and the U.S. Preventive Services Task Force recommend at-risk adults undergo low-dose CT screenings annually, lung cancer is often detected when it has reached more advanced stages that are harder to treat.
By searching where cancer is likely to occur, Sybil tells doctors where to look, increasing the potential for spotting signs early when cancer is more treatable and with it, survival rates.

“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,” said Dr. Lecia Sequist, program director of the Cancer Early Detection and Diagnostics Clinic and a lung cancer medical oncologist at Mass General Cancer Center, in a statement.

Despite previous studies showing that low-dose CT reduced risk of death from lung cancer by 24%, less than 10% of eligible patients get annual screenings.

Sybil looks for signs of abnormal growths in lungs, but it’s the patterns and nuances that it evaluates within these scans that scientists cannot, that make the technology a predictive tool rather than just a detection solution.

The scientists say that it is able to predict a person’s chances of developing lung cancer between one and six years in advance. In fact, in some cases, it has detected signs of cancer that radiologists could not even identify until nodules were visible on a CT scan years later.

Despite its potential, the researchers told NBC News that the data used to train it does not include “sufficient Black or Hispanic patients to have confidence in broad applicability yet.”

The FDA announced last year that researchers and companies must have plans for ensuring clinical trials represent diverse patient populations to test medical products before they can be submitted for approval.

Overdiagnosis is also a concern, with doctors looking to avoid unnecessary biopsies for benign tumors.

Patients and providers should not misinterpret the system as a replacement for radiologists either.

“The future of radiology is going to be AI-assisted. You will still need a radiologist to identify where the cancer is, identify the best possible treatment and actually do the treatment,” Dr. Florian Fintelmann, a radiologist at Mass General Cancer Center and one of the researchers working on Sybil, told NBC News.

CT Homepage