AI system accurately interprets cardiac MRI scans
Experts have developed an artificial intelligence-powered system they say can significantly improve cardiac imaging analysis.
Cardiac MRIs are known to be among some of the most challenging imaging exams to interpret; the process is time-intensive, even for the most seasoned radiologists. As such, simplifying these interpretations has been the target of experts for many years.
Developed by a team of researchers from Carnegie Mellon University with help from Cleveland Clinic’s Cardiovascular Innovation Research Center, experts are hopeful the system—called "CMR-CLIP"—can support radiologists tasked with interpreting cardiac MRIs. The vision language model combines moving images from the scans with patients’ associated reports.
“Cardiac MRI interpretation is highly specialized and time intensive,” said co-principal investigator of the analysis David Chen, PhD, with the Cleveland Clinic. “Systems like CMR-CLIP have the potential to support clinicians through automated screening, and interpretation support, particularly in settings where expert readers are limited. Such reader assistant tools are critical to improving patient access to this powerful diagnostic technology.”
CMR-CLIP (cardiovascular magnetic resonance imaging-contrastive language image pretraining) was trained on over 11,000 cardiac MRI exams and their corresponding reports. This enabled researchers to develop a capable model without the time-consuming task of manually annotating training data. Through this, the model was able to learn how physicians describe these exams in clinical settings.
Testing revealed the system to be capable at clinical tasks, achieving accuracy of 88.5% for identifying non-ischemic cardiomyopathy, 88% for ischemic cardiomyopathy, 96.2% for cardiac amyloidosis and 98.6% for hypertrophic cardiomyopathy. It was able to identify abnormalities in a zero-shot setting as well, meaning that it had not been previously trained on that specific finding.
What’s more, the system also performed better than other AI models developed for the same purpose. It was up to 35% more accurate in some cases, even when trained on just a single example of a specific imaging feature. This performance was maintained on exams acquired at outside institutions as well.
“This work highlights a new direction for medical AI by showing how large-scale clinical data can be used to train models without requiring time-consuming manual labeling,” noted Deborah Kwon, MD, director of Cardiac MRI at Cleveland Clinic, a clinical lead and co-author of the study. “This technology has the potential to not only improve efficiency but also quality of reporting to support more consistent and clinically meaningful interpretations, as well as serve as an important teaching tool in a highly specialized and complex imaging field.”
Read more about the team’s work in Nature Communications.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.