Combining AI with OCT shows potential for detecting lipid-rich plaques in coronary arteries: Study

Researchers have developed a new artificial intelligence-based approach for detecting fatty deposits inside coronary arteries using optical coherence tomography (OCT) images. Because these lipid-rich plaques are strongly linked to serious cardiac events such as heart attacks, the method could eventually help doctors spot dangerous plaques before they rupture and cause damage.

OCT is used during catheter-based procedures such as those used to open partially blocked blood vessels and place stents to help blood flow more freely. Although OCT provides very detailed images of the vessel structure, standard OCT images don’t reveal the composition of the vessel wall, which is important for assessing heart attack risk.

“Plaques with more lipid and certain patterns of lipid distribution are strongly associated with the risk of major cardiac events,” said research team leader Hyeong Soo Nam from Korea Advanced Institute of Science and Technology in South Korea. “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.”

In the Optica Publishing Group journal Biomedical Optics Express, the researchers describe their new method for extracting spectral information from OCT images. They also developed a deep learning approach that enables quantitative, automatic assessment of lipids directly from intravascular OCT images. The new method doesn’t require any hardware changes and works with OCT systems already used in the clinic.

“During a coronary intervention, this method could provide clinicians with additional information to support risk assessment, procedural planning and evaluation of treatment response,” said Nam. “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.”

Extracting spectral information

Although OCT is used in clinical practice, identifying lipid-rich, high-risk plaques still depends heavily on the physician’s experience. For several years, the researchers have been working with Jin Won Kim’s team at Korea University Guro Hospital to overcome the limitations of conventional OCT.

“Our group previously demonstrated that spectroscopic OCT can detect lipid-related optical signatures within atherosclerotic plaques,” said Nam. “This new study builds on that by extending it with modern deep learning techniques to significantly improve detection accuracy and robustness.”

The new method feeds wavelength-dependent information from OCT images into an AI model. This is possible because different types of tissue interact with light in different ways. Lipid, fibrous tissue and calcium, for example, each absorb and reflect light in slightly different ways. The AI model learns to recognize signal patterns that are more likely to originate from lipid-rich tissue and can then automatically highlight suspicious regions throughout the image.

“Importantly, unlike many conventional AI systems that require experts to painstakingly label lipid regions at the pixel level — an extremely time-consuming and subjective process-our approach learns from much simpler frame-level annotations that indicate only whether lipid is present or absent,” said Nam. “This substantially lowers the annotation burden and makes the method far more practical for real-world clinical use.”

AI predictions vs histology

The researchers validated their new approach by using intravascular imaging data acquired from a rabbit model of atherosclerosis. They compared the AI-generated predictions against histopathology results obtained using lipid-specific tissue staining, evaluating how accurately the method identified image frames containing lipid-rich plaques and whether it highlighted anatomically meaningful regions.

“The results showed strong classification performance along with good spatial agreement with the pathological findings,” said Nam. “Looking ahead, the same framework we applied could be extended to other intravascular or optical imaging modalities where subtle spectral or signal variations are present but underutilized.”

The researchers are now working to improve the processing speed and robustness of the approach to make it more practical for real-time clinical use. They also plan to perform additional validation studies using human coronary artery data and figure out the best way to integrate the method into existing clinical workflows in a way that is seamless for physicians.

Reference:

Jin Hwan Hwang, Woojin Lee, Jin Hyuk Kim, Ryeong Hyun Kim, Dong Oh Kang, Jin Won Kim, Hongki Yoo, and Hyeong Soo Nam, Automated lipid detection in spectroscopic optical coherence tomography using a weakly supervised deep learning network, Biomedical Optics Express, https://doi.org/10.1364/BOE.585222