A model that calculates the probability of cancer may help to rule out a lung cancer diagnosis following screening preventions – which would substantially reduce false positives.
A model that calculates the probability of cancer may help to rule out a lung cancer diagnosis following screening preventions — which would substantially reduce false positives
By Katie Kosko
A model that calculates the probability of cancer may help to rule out a lung cancer diagnosis following screening preventions — which would substantially reduce false positives, according to study findings published in Thorax.
The process involves artificial intelligence (AI), or intelligence demonstrated by machines that learn from experience, such as self-driving cars.
People who are considered at high-risk for lung cancer are generally screened for the disease using low-dose computerized tomography (CT) scans. But, oftentimes, these provide false positives. “About a quarter of the screened at-risk population presents some nodule or shadow in the lung,” Panayiotis (Takis) Benos, Ph.D., professor and vice chair of computational and systems biology, and associate director of the Integrative Systems Biology Program at the University of Pittsburgh, said in an interview with CURE. “Thankfully, 96 percent of them are benign, hence the term ‘false positives.’ However, in order to avoid misdiagnosing a real case of cancer, we need to follow these cases with additional screenings and/or biopsies, which increase anxiety, health risks and health care costs.”
The researchers used CT scan data from 218 patients at the University of Pittsburgh Medical Center who either had confirmed lung cancer or benign nodules. Then they used AI technology to create a model that can calculate the probability of cancer. If the probability falls below a certain threshold, the model rules out cancer, explained the researchers.
The researchers found that they would have been able to save 30 percent of the people with benign nodules from undergoing additional testing, without missing a single case of cancer.
“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,” senior author David Wilson, M.D., M.P.H., associate professor of medicine, cardiothoracic surgery and clinical and translational science at the University of Pittsburgh and co-director of the Lung Cancer Center at UPMC Hillman, said in a press release.
The model examined three important factors: the number of blood vessels surrounding the nodule, the number of nodules and the number of years since the patient quit smoking. “In order for cancers to grow, they trigger development of nourishing blood vessels to deliver growth factors and proteins to help the cancer grow and spread,” said Wilson. “This process does not occur in non-cancerous nodules. If we can develop a way to reliably measure the number of blood vessels in a nodule, it is logical to assume that the higher the vessel number, the more likely a nodule is cancer.”
Researchers have already begun to examine this AI model using 6,000 scans from the National Lung Screening Trial. “AI technology has recently shown promising results in many aspects of clinical practice,” said Benos. “But we need to proceed with caution and make sure we understand the components of each model and its limitations.”
People who are eligible for lung cancer screening with CT scan should still consider having it done, said Wilson. “We are continuously working on ways to improve the interpretation and meaning of the scan results,” he added.