Development of a Semi-Automatic Tool for Coronary Artery Plaque Segmentation and Classification from OCT Images
Supervisor: Dr Sathish Kumar Marimuthu
School: Engineering
Description:
Coronary arteries are the blood vessels that supply blood to the heart muscle. They can become narrowed by plaque, reducing blood flow and potentially leading to heart attacks. Optical Coherence Tomography (OCT) is an intravascular imaging technique in which a catheter is inserted into diseased coronary arteries to obtain high-resolution images of the vessel wall. OCT images provide interventional cardiologists with information about the degree of narrowing and the type of plaque present. Based on this information, cardiologists decide the type and size of stent to implant to restore blood flow. In our research, we use OCT derived 3D coronary artery geometries to simulate percutaneous coronary intervention (PCI) procedures, with the long-term aim of improving treatment planning and procedure optimisation.
This summer project will develop tools to extend an existing OCT image segmentation workflow. The current code can segment the lumen boundary (innermost layer of the blood vessel) from OCT data, but further development is needed to identify additional vessel-wall features, including plaque constituents and the external elastic lamina. The student will work towards developing a semi-automatic image analysis tool for coronary plaque segmentation and classification, producing outputs that can support future patient specific PCI simulations. This will help address an important challenge in patient-specific PCI modelling, as each patient has a unique coronary anatomy and degree of stenosis. These differences affect how a stent expands inside the diseased artery and may influence the risk of restenosis or future reintervention.
The student will test image-processing methods such as filtering, edge detection, boundary detection, and segmentation overlays on selected OCT frames. The project will be designed with staged deliverables so that useful progress can be made within the 8 -10 week internship, even if fully automated segmentation is not achieved.