AI To Make Cardiovascular Imaging Smarter

0

Yu Gan, assistant professor in the Department of Biomedical Engineering at Stevens Institute of Technology, traces his fascination with artificial intelligence back to the computer games of his childhood.

“In Chinese, the word ‘computer’ means ‘electrical brain,’ and I wondered if a computer could be as smart as a brain,” recalled Gan. “When I got my first computer, an old one with Windows 98, I played computer games such as Heroes of Might and Magic, and I realized that a computer can be designed to be smart — smarter, at least, than the boy who was playing with it!”

Then, the mysteries of diagnostic technology sparked Gan’s curiosity about biomedical imaging.

“My high school physics teacher showed us the first X-ray radiograph image to demonstrate how advancements in physics led to a revolutionary method in helping doctors almost magically see through the human body,” Gan said. “Closer to home, my maternal grandmother suffered from coronary disease for many years. When my mom would report that doctors advised that [my grandmother] didn’t need surgery yet to enlarge the narrowed blood vessels, I wondered how the doctor could see the vessels and what he actually saw.”

Today, Gan has combined those interests into his pioneering research to apply artificial intelligence to reveal hidden information in images from optical coherence tomography (OCT), a noninvasive diagnostic imaging technology.

“We’re developing robust algorithms including artificial intelligence to both process complex images and make those images more informative and meaningful,” he explained. “The goal is to aid the diagnosis, clinical decision-making and treatment for cardiovascular disease, the leading cause of death and disability for people worldwide. There is a very real need to address these critical clinical issues.”

At the imaging quality level, Gan’s research aims to improve resolution and reduce noise. At the perception level, he is training the algorithms to automatically identify disease-related patterns and regions.

“We are not just presenting higher-quality images more quickly,” he said. “We also want to learn something from the images. We are working to turn grayscale images into colorful images with more disease-related information to unveil the underlying patterns. As a bonus, this method costs much less than traditional histology. Just 10 years ago, this visualization method was beyond imagination, but we can do it now, thanks to algorithms that teach the computer to create new kinds of imagery better suited to biomedical analysis.”

Integrating his research with his teaching, Gan launched a graduate-level biomedical engineering course, “Machine Learning in Biomedical Engineering,” for the Spring 2023 semester. Students explore examples of the power of machine learning and artificial intelligence on biomedical engineering applications.

“I’m doing biomedical image processing research to improve human health,” Gan said. “I bring that research to this course to give my students a better sense about AI and biomedical engineering, and I collect feedback and ideas from the students to positively impact my research.”

Gan is also testing his research with Brigitta Brott, a cardiologist, and Silvio Litovsky, a pathologist, at the University of Alabama at Birmingham’s Heersink School of Medicine. Future plans include moving from bench to bedside to extend his groundbreaking work into a more clinical setting and expanding to collaborate with additional hospitals.

Gan’s coronary imaging research work is supported by funding from the National Science Foundation and the New Jersey Health Foundation. He has also extended his AI research to investigate additional uses, including wearable sensor images to study children’s behavior, through a National Institute of Health grant; confocal laser scanning for cancer-related studies, supported by the Burroughs Wellcome Fund; and millimeter wave image processing to advance food science, funded by the U.S. Department of Agriculture.

While he is still captivated by the challenge of making computers smarter in the field of biomedical imagery, Gan’s real satisfaction comes from his work’s life-changing potential.

“It makes me proud to be working on the leading cause of human deaths,” he said. “Improving the quality of coronary imaging can help cardiologists make better decisions — for my grandma and for all patients who deal with similar situations.”