This is one of several deep learning algorithms developed by Mass General and Brigham and Women’s to look at standard chest X ray To use artificial intelligence to perform a variety of opportunistic examinations of patients who undergo chest x-rays for various reasons. They developed algorithms to look at chest X-rays for lung scans, determine a patient’s biological age to determine longevity better than current methods, and now cardiac assessments. Artificial intelligence is still in development, but it represents a new trend in radiology, which will likely see it commercialized by several vendors within the next couple of years.
Since chest X-rays are the most common radiological examination performed worldwide, AI development efforts are focused on this type of test. Artificial intelligence extracts thousands of tiny radioactive data points that humans cannot see.
“We found that there was very strong information captured by chest X-rays that we weren’t aware of and couldn’t identify, but with deep learning techniques we can now extract that information and use it to predict risk,” Weiss said. .
Artificial intelligence can predict the 10-year risk of dying from a heart attack or stroke. The algorithm was trained using 150,000 chest radiographs from 40,643 patients, in which patient history and outcomes were known. The deep learning system was able to look at patient data and X-rays and identify radiological patterns in the images that were the same in patients with similar outcomes. It was then tested on about 11,000 radiographs from other patients without any additional information and the AI was able to accurately determine the risk of these patients using the imaging data alone.
In radiology, it’s not a thing or two as simple as seeing an enlarged heart that becomes apparent to a radiologist, Weiss said. It’s a combination of thousands of tiny data points in an image that an AI can quickly evaluate but a human would miss. Research is not 100% sure what the AI is looking for, he said, but the self-learning algorithm produced accurate results when it was applied to new patients.
That actually gets to the heart of this study. Essentially [the AI] It is the “black box”, because we just feed the X-rays into the network and it broadcasts a risk prediction about how likely the patient is to experience some cardiovascular event in the future. Weiss explained that we can’t identify which features are responsible for the prediction. “We can’t say which anatomical changes or other changes in the image actually contribute to the final prediction.”
This kind of AI “black box” technology where AI finds complex patterns based on thousands or hundreds of thousands of small image data points, may be the way of the future for all kinds of AI-based risk prediction algorithms. If accepted by the Food and Drug Administration, these types of AI would evaluate all X-rays taken in a hospital or clinic to automatically screen patients in the background and add additional value to each test. This may aid efforts in early detection and prevention of the disease before patients become severely ill.
Weiss said American College of Cardiology (ACC) and the American Heart Association (AHA) Both recommend risk assessments for patients to aid in primary prevention efforts of cardiovascular disease. He said those assessments are based on nine forecast data points.
“We compared our predictions of cardiovascular risk based on chest X-rays with this recommended risk prediction guideline, and what we found in this test group of 11,000 patients is that we are not significantly worse than the recommended risk predictors,” he said. “Overall, we are as accurate as current risk predictions are.”