How Artificial Intelligence Is Helping Doctors Detect Heart and Lung Disease
How Artificial Intelligence Is Helping Doctors Detect Lung and Heart Disease
Radiologist Seth Kligerman, MD, pulls up a series of scans on his monitor — blobs of black and gray to the untrained eye. “These are CT scans of lungs. We’re looking for nodules,” he says, noting that each scan provides roughly 700 image slides of the insides of lungs. In the past, doctors would have to review scans manually, squinting over low-resolution images for signs of potentially cancerous nodules, which can be as small as one to two millimeters in diameter. “It’s easy for even experienced doctors to miss lung nodules,” says Dr. Kligerman. “And this is where AI-enhanced technology can be a big help.”Dr. Kligerman and colleagues at National Jewish Health are using algorithms that have been scientifically vetted and approved by the U.S. Food and Drug Administration to help assess scan results. These tools pore over millions of data points to identify patterns consistent with health risks, transforming hours of work for human doctors into minutes of automated output.
The technology has the potential to improve diagnosis for diseases that affect different parts of the body, making it ideal for multispecialty hospitals. Dr. Kligerman moves on from lung scans to images of the heart, demonstrating how the algorithms have also been helping doctors identify things like stenosis, where a buildup of calcium can cause the valves to narrow, increasing risks of heart attack and stroke. He points to the white spots where the AI has marked calcification. “That’s at 40%,” he says, making a note. “The readouts are very detailed. Sometimes, they can be a little too sensitive, so we need to review them ourselves before coming to a conclusion. It also helps that we have these new CT scanners that can give us clearer, more detailed images.”
Using AI to Detect Lung Disease Earlier
The AI analysis tools being used by Dr. Kligerman and others have been specially designed for medical professionals. “There’s a distinction between the AI tools that we develop or use in-house — when we’re looking to measure fibrosis (scarring), for instance — and the tools people have on their desktop,” says researcher Stephen Humphries, PhD.
Dr. Humphries, whose background is in physics and biomedical engineering, has spent years developing and validating AI analysis models for lung images. “We’ve developed tools that can detect and quantify features in CT scans,” he explains. “Increasingly, we’re comparing those features with genetic information or blood-based biomarkers. So, both in terms of clinical care and research, there’s a lot of possibility.”
One of those possibilities lies in detecting disease earlier and tracking its progression with greater accuracy. “The amount of fibrosis, or scarring, in the lung — what percentage of the lung appears to be affected — is a strong indicator for disease,” Humphries says. “This can help inform treatment decisions. We can also detect things that were previously missed.” The AI system Dr. Humphries helped develop, a data-driven textural analysis method (called DTA) can analyze high-resolution CT scans to measure scarring and track changes over time. It can even help predict how those changes might affect a patient’s health.
For physicians treating interstitial lung disease (ILD), this kind of insight is invaluable. “The extent of scarring in the lungs, as detected by quantitative CT analysis, has been shown to be predictive of outcomes,” says Joey Pryor, MD, who recently led a study on the subject published in the American Journal of Respiratory and Critical Care Medicine.
Dr. Pryor explains that these algorithms can pick up subtle changes that the human eye may not easily detect. “We’re looking at such fine detail when we’re comparing CT scans that it’s difficult to visually pick up on these things,” he says. “So using the quantitative CT scans and AI analysis definitely gives us an advantage.”
The Future of AI and Health Care
When it comes to ILD, Dr. Pryor and his colleagues are using AI tools primarily in research. However, the goal is to make them part of routine screenings. “The hope is that in the next couple of years this stuff is incorporated into all CT scans for ILD,” says Dr. Pryor.
Even so, doctors are approaching these innovations with a healthy dose of skepticism. “One concern is that all this is happening faster than we’re able to totally validate things,” Dr. Humphries cautions. “But I think we’re positioned as an institution to be rigorous in validation and double-checking.”
“Having a radiologist is by no means being replaced,” Dr. Pryor adds. “An AI tool provides an objective variable that can be added to our tool belt in assessing disease progression.”
For patients, the promise of AI is not in the novelty of the technology but in how it enhances care. “Right now, we track disease progression based on symptoms, breathing tests and radiologists’ interpretations,” says Dr. Pryor. “It’s possible that quantitative analysis is more sensitive for picking up smaller changes. It can give a more precise measurement — like that you went from 5% scarring to 7% scarring — instead of just saying there’s been mild progression.”
Combining these AI instruments with human expertise could reshape the standard of care for complex lung and heart diseases. “It’s really about how we use the information,” says Dr. Humphries. It’s about taking data from imaging, diagnostics and clinical observations and turning it into insights that can help people live longer, healthier lives.”