Stented coronary arteries: a semi-automatic segmentation method for OCT-based reconstruction

The influence of coronary anatomy on percutaneous coronary intervention (PCI) outcomes is largely confirmed in the literature [1],[2]. Indeed, there is no consensus about the optimal stenting technique for PCI due to the significant anatomic variability observed between individuals [2]. In this context, the application of medical imaging techniques to reconstruct patient-specific stented coronary arteries beside the application of computational tools, such as computational fluid dynamics (CFD), can be crucial to fulfill clinical demands. Nowadays, optical coherence tomography (OCT) is widely adopted during PCI [1]. Information provided by OCT can be used to emulate the coronary anatomy that characterizes each individual. Particularly, the high resolution ensured by OCT acquisitions (i.e. 10 - 20 μm) enables to obtain an exhaustive picture of the blood vessel lumen both before and after the stent positioning [2]. The major drawback is related to the fact that OCT images are lacking of information on the three-dimensional (3D) shape of the coronary tree. As a consequence, to reconstruct the coronary 3D geometry, other diagnostic techniques, such as angiography or computed tomography, can be used to extract the vessel centreline [3], along which OCT slices can be aligned. This study proposes a method to reconstruct 3D stented coronary vessels by means of OCT images and its validation. The method was applied to the OCT scan of a rigid phantom emulating a typical coronary left anterior descending (LAD) segment with bifurcations. The method is characterized by automated lumen and stent segmentation of the OCT images and their alignment with the vessel centreline, which was obtained from a X-ray computed micro-tomography (μCT) of the phantom.