There are more than 300 FDA-approved AI tools for use in radiology, according to Anant Madabhushi, a professor in the department of biomedical engineering at Emory University School of Medicine in Atlanta. Most are used to help doctors diagnose and treat cancer, he said, but not to predict someone’s future cancer risk., like Sybil does.
Sybil looks for signs of where the cancer is likely to appear, so that doctors know where to look and can detect it as soon as possible.
Lung cancer is most treatable when it’s detected early, said Dr. Kim Sandler, an associate professor of radiology at Vanderbilt University Medical Center in Nashville, Tennessee.
But early detection is difficult, he said. Since the lungs cannot be seen or felt, the only way to detect it early is with a CT scan. By the time symptoms appear, such as a persistent cough or shortness of breath, the cancer is often advanced and more difficult to treat.
Past investigation has shown that low-dose CT screening can reduce the risk of death from lung cancer by 24%, because they can help detect cancer earlier, when it is most treatable.
But an artificial intelligence tool could potentially increase early detection rates for lung cancer, and also potentially increase survival rates, Sandler said.
“This is great for radiologists, but more importantly, it’s really great for our patients to be able to give them that tool to tell them whether or not we think something is lung cancer,” he said.
How does Sybil work?
To predict cancer risk, Sybil relies on a single CT scan. It scans the three-dimensional image, looking not only for signs of abnormal growth in the lungs, but also for other patterns or complaints that scientists don’t yet fully understand, said Dr. Florian Fintelmann, a radiologist at the Mass General Cancer Center and one of the researchers working on Sybil.
Based on what she sees, Sybil provides predictions about whether a person will develop lung cancer in the next one to six years, she said.
There have been cases, Fintelmann added, in which Sybil has detected signs of cancer that radiologists missed until the nodules were visible on a CT scan years later.
Fintelmann said he sees a future where the AI tool is helping radiologists make important treatment decisions, without replacing radiologists entirely.
“The future of radiology will be assisted by AI,” he said. «You’ll still need a radiologist to identify where the cancer is, identify the best possible treatment, and actually do the treatment.»
The AI, however, is still far from perfect.
One issue Madabhushi of Emory University said worries him is the type of data used to train the AI.
“A lot of data that comes from medical institutions or clinical trials does not represent the diversity of our country,” he said, adding that he believes artificial intelligence tools are not being developed in a way that is designed to help black and brown people. .
The scientists who developed Sybil have admitted that the data used to create the AI tool does not include «enough black or Hispanic patients to have confidence in broad applicability yet.»
The FDA has already taken a step to address this issue, announcing last year which will soon require researchers and companies seeking approval of medical products to submit a plan that ensures diversity in clinical trials.
“We have to make sure that the AI does not reflect or capture our biases,” Madabhushi said.
There are also concerns about overdiagnosis, Sandler said. Doctors may subject patients to a potentially unnecessary biopsy for a nodule that may turn out to be benign.
“Are we finding a cancer that maybe we don’t need to find?” she said.