The revolution is here: AI’s growing role in cardiovascular imaging, interventional cardiology

Artificial intelligence (AI) has made a massive impact on healthcare, especially when it comes to cardiology and radiology, so it should come as no surprise that AI is being fully embraced by the fields of cardiovascular imaging and interventional cardiology. 

A new analysis in JSCAI, the official journal of the Society for Cardiovascular Angiography and Interventions (SCAI), examined this trend in great detail.

“Facilitating complex interventional procedures requires accurate image acquisition and interpretation as well as precise decision-making,” wrote co-authors Maryam Alsharqi, DPhil, and Elazer R. Edelman, MD, PhD, with the Massachusetts Institute of Technology in Cambridge. “However, the traditional practice of image acquisition, analysis and decision-making presents issues in timing, scaling and diagnostic accuracy. AI and machine learning applications in cardiovascular imaging are rapidly gaining prominence as generated images become increasingly complex. AI algorithms have the capacity to streamline the clinical workflow through rapid and accurate image interpretation and quality control. They can help expand our understanding of certain interventions to improve clinical decision-making, particularly with the recent advances in stent technology and transcatheter valve replacement.”

More and more AI applications are being cleared by the U.S. Food and Drug Administration (FDA) to be used in cardiology and radiology—click here for the latest numbers for each specialty—and that momentum is only going to grow stronger as time goes on.

Alsharqi and Edelman noted that many of these FDA-cleared AI applications are being built for the worlds of noninvasive cardiovascular imaging and interventional cardiology. 

The 4 primary AI techniques found in cardiac imaging

1. Image processing and generation

AI’s ability to enhance images and even generate new images of its own has been a game-changer in the field of cardiac imaging. It helps clinicians overcome some of the many challenges associated with medical imaging—long acquisition times and radiation exposure, for example—and results in significant improvements in the quality of care heart patients receive. 

“Under sampling is one technique that has significantly accelerated cardiac MRI scan times by acquiring fewer images and estimating the remainder,” the authors wrote. “Computer vision for virtual image creation is another approach to lower radiation dose by reducing the number of acquired imaging series. It allows contrast-enhanced CT images to be created from non-enhanced images and vice versa. Generative adversarial networks have the capacity to synthesize cine-like cardiac MRI from real-time sequences, helping patients with arrhythmia or those who struggle during breath-holding image acquisition, resulting in improved image quality.”

2. Image classification and segmentation

Some physicians feared early on that AI may replace them in the cath lab. Now, however, we see that AI’s role is to save time so that physicians can spend more of their days communicating with patients and performing complex procedures. 

“Researchers have recently shown that using a multiscale deep reinforcement learning approach allowed accurate anatomical landmark identification of 5,000 3D cardiac CT images (2.5 million slices) in less than a second,” Alsharqi and Edelman wrote. 

The authors also emphasized the progress being seen with AI-powered image segmentation, which is yet another way these algorithms can save clinicians valuable time.

3. Association and prediction

“As the computing power of AI has the capacity to handle big and complex data sets, discovering statistical associations, underlying physiological interactions, and meaningful clinical insights has become more feasible,” the co-authors wrote. “Association analysis methods, such as support vector machines and random forests, facilitate the promise of precision cardiovascular medicine through the identification of patient-specific clinically relevant associations by integrating clinical symptoms, imaging data and various biomarkers.”

AI is already helping identify suspicious signs of cardiovascular disease in echocardiography, coronary CT angiography and other modalities. Just one walk through the exhibition hall at any major healthcare conference will make it clear just how many medtech companies are focused on this specific aspect of AI-powered cardiovascular imaging.

4. Decision support

AI-enabled clinical decision support (CDS) systems can track electronic health record data and other hospital records for key information that can then be used to ensure patients are receiving the best care possible. 

“With the advancement of AI techniques, particularly in natural language processing, clinical decision support systems have become more flexible and can be customized to fit the needs of patients and clinical care teams,” the group wrote.