Researchers have demonstrated how artificial intelligence can enhance optical coherence tomography (OCT) to better identify lipid-rich plaques in coronary arteries. An important step toward earlier detection of high-risk lesions that can lead to heart attacks.
OCT is widely used during catheter-based coronary interventions, such as stent placement, because it provides high-resolution images of vessel structures. However, conventional OCT offers limited insight into the composition of the vessel wall, while plaque composition, particularly lipid content, is a key indicator of cardiovascular risk.
“Plaques with more lipid and certain patterns of lipid distribution are strongly associated with the risk of major cardiac events,” explains Hyeong Soo Nam, research team leader at the Korea Advanced Institute of Science and Technology. “By analyzing wavelength-dependent information hidden in the OCT signal and combining it with AI, we were able to identify the presence and distribution of lipid within the vessel wall.”
Spectral information from OCT images
In a study published in Biomedical Optics Express, the researchers describe a method that extracts spectral information from standard OCT images and feeds it into a deep-learning model. The result is an automated, quantitative assessment of lipid distribution directly from intravascular OCT scans, without requiring any hardware modifications. This makes the approach compatible with OCT systems already in clinical use.
According to Nam, the technology could add valuable context during coronary interventions. “During a coronary intervention, this method could provide clinicians with additional information to support risk assessment, procedural planning and evaluation of treatment response,” he says. “Ultimately, it has the potential to contribute to safer clinical decision making, more individualized treatment strategies and improved long-term management of patients with coronary artery disease.”
Beyond visual interpretation
At present, identifying lipid-rich plaques on OCT images largely depends on the clinician’s experience. To address this limitation, the research team has collaborated for several years with clinicians at Korea University Guro Hospital. Earlier work showed that spectroscopic OCT can reveal lipid-related optical signatures within atherosclerotic plaques. “This new study builds on that by extending it with modern deep learning techniques to significantly improve detection accuracy and robustness,” Nam says.
Different tissue types, such as lipid, fibrous tissue and calcium, interact with light in distinct ways. By learning these subtle signal patterns, the AI model can automatically flag regions likely to contain lipid-rich plaques. Importantly, the approach reduces the need for labor-intensive pixel-level annotations.
“Unlike many conventional AI systems that require experts to painstakingly label lipid regions at the pixel level, our approach learns from much simpler frame-level annotations that indicate only whether lipid is present or absent,” Nam notes. “This substantially lowers the annotation burden and makes the method far more practical for real-world clinical use.”
Validation and next steps
The method was validated using intravascular imaging data from a rabbit model of atherosclerosis, with AI predictions compared against lipid-specific histological staining. The results showed strong agreement between the AI output and pathological findings, both in classification accuracy and spatial localization.
The researchers are now focused on improving processing speed and robustness to support real-time clinical use. Additional validation with human coronary artery data is planned, alongside efforts to integrate the technology seamlessly into existing clinical workflows.
