- Researchers at the University of Pittsburgh and UPMC Hillman Cancer Center are using machine learning to substantially reduce false positives in lung cancer screenings.
Lung cancer is the leading cause of cancer deaths worldwide, the researchers noted, and low-dose CT screenings are critical for early detection and diagnosis. Nationwide, about a quarter of these scans show shadows that indicate nodules in the lung, which is considered a positive result.
However, fewer than four percent of these individuals truly have cancer, meaning this method has a 96 percent false positive rate.
Based on the imaging test alone, human physicians cannot determine which of the nodules are benign and which indicate cancer. As a result, patients often undergo invasive and expensive testing to determine the nature of the nodules.
“A positive test creates anxiety, increases health care costs, and the follow-up tests are not risk free,” said Panayiotis (Takis) Benos, PhD, professor and vice chair of computational and systems biology and associate director of the Integrative Systems Biology Program at Pitt.
“For the 96 percent of people who have benign nodules, these procedures are unnecessary. So, we try to mine the data to tell which are benign and which are malignant.”
The research team collected low-dose CT scan data from 218 high-risk UPMC patients who were later confirmed to have either lung cancer or benign nodules.
The group then developed a machine learning algorithm that would calculate the probability of cancer. The model examined several factors, including the number of blood vessels surrounding the nodule, the number of nodules, and the number of years since the patient quit smoking. If the probability falls below a certain threshold, the algorithm rules out cancer.
When comparing the model’s conclusions to the actual diagnoses, the team found that they would have been able to prevent additional testing for about 30 percent of individuals with identified nodules.
“We were able to rule out cancer in about a third of patients, so they wouldn’t need biopsies, they wouldn’t need PET scans or a short-interval CT scan. They just need to come back in a year,” said senior author David Wilson, MD, MPH, associate professor of medicine, cardiothoracic surgery and clinical and translational science at Pitt and co-director of the Lung Cancer Center at UPMC Hillman.
Machine learning and artificial intelligence algorithms have shown potential in reducing false positive rates for several different types cancer screenings.
In February 2018, a team from Johns Hopkins School of Medicine developed a test supported by machine learning that accurately detected the presence and origin of several different cancers, offering the opportunity for less invasive and more precise testing.
The UPMC study is the first time researchers have used artificial intelligence to distinguish between benign and cancerous nodules in lung cancer screening. The team plans to continue their work and apply the algorithm to larger datasets.
“While it has been known for some time that tumors recruit more vascular support, this is the first time that we’ve been able to use computer technology to quantify their contribution and incorporate them into a predictive model that decides, with certainty, that some patients don’t have cancer,” Wilson said.
“The next step is to evaluate this technique in a larger population, using about 6,000 scans from the National Lung Screening Trial.”