Artificial intelligence (AI) may one day improve the accuracy of computed tomography (CT) scans for lung cancer screening, according to a 2019 study in the journal Thorax.
Although 96 percent of the suspicious masses that CT scans detect turn out to be benign, all CT-detected masses must be investigated further, either with more imaging tests using radiation or potentially dangerous lung biopsies. Reducing the number of “false-positive” CT scans would spare patients unnecessary—and anxiety-provoking—procedures.
Researchers collected data such as smoking history and the nature and number of lung masses for 218 patients who had been screened for lung cancer and were subsequently confirmed either to have malignant lung tumors or to be cancer-free. Using AI techniques, they found that three variables helped to differentiate malignant nodules from benign ones: the number of years since a patient had quit smoking, the total number of nodules in his or her lungs, and the number of blood vessels feeding a nodule.
The researchers calculated that they would have been able to prevent 30 percent of patients with benign nodules from undergoing further testing, without missing any cases of cancer. These findings are limited by the small sample size and the fact that all participants were at high risk for lung cancer.
If these results are confirmed in larger clinical trials, AI technology could reduce the number of false-positives in lung cancer screenings.