Artificial intelligence (AI) could be used to improve the prediction of stent expansion in severely calcified lesions. A presentation given Monday at Cardiovascular Research Technologies (CRT) 2024 in Washington, D.C., describes the development and validation of an established intravascular ultrasound (IVUS)-derived calcium score using AI-based automation to predict stent expansion potential in lesions with significant calcium content. “This research has the potential to transform the way we approach stent placement in severely calcified lesions," said Saurav Chatterjee, MD, clinical assistant professor of medicine at the Donald and Barbara Zucker School of Medicine at Hofstra/Northwell in New York. "By harnessing the power of AI and IVUS imaging, we aim to provide clinicians with a reliable and reproducible tool for predicting stent expansion, potentially improving patient outcomes in the long run.” Measuring calcium arc The pilot study proposes the use of supervised learning to automatically measure the arc of calcium located on the coronary vessel wall that is detected using IVUS images. The results will then be used to build an AI-based automated IVUS-derived calcium score to assess the potential for suboptimal stent expansion on pre-intervention. Chatterjee, who presented at the Late Breaking Artificial Intelligence Clinical Trials session, explained that the use of unsupervised learning could identify new features associated with stent underexpansion and match optimal tools needed for achieving improved stent expansion. The research aims to address the observation that identifies stent underexpansion in coronary arteries as a cause of stent thrombosis (ST) and restenosis. The strong association with (largely) superficial calcium extent and thickness further compounds this challenge. “Intravascular imaging is associated with lower risk of ST,” said Chatterjee. “IVUS and optical coherence tomography (OCT) can both assess extent of vessel wall calcification. “IVUS is used more in the real-world and does not require contrast, can use a manual and automated pullback, and is of heterogenous quality.” Study outcomes The research is supported by a team of co-investigators that include Sahil Parikh, MD, and Partha Sardar, MD, both from Columbia University Irving Medical Center in New York, and Ronny Shalev, PhD, from Dyad Medical in Boston. Commenting on the study’s possible outcomes, Chatterjee said that the anticipated results could be comparable to a human annotator. He acknowledged possible research limitations that were dependent on the quality of acquired images that used a number of resources for training purposes. There was also the element of time invested in the research and the degree of interruption to workflows, while the AI technology was being developed. “[The existence of] tools and established methods indicate reasonable probability of successful implementation,” said Chatterjee. “The research represents a collaborative effort to advance cardiovascular care through innovative technology and research methodologies.” Photo Credit: Jason Wermers/CRTonline.org Photo Caption: Saurav Chatterjee, MD, describes an intravascular ultrasound-based artificial intelligence system to determine calcification Monday during CRT 2024 in Washington, D.C.