Automatic stent segmentation in IOCT images using combined feature extraction techniques and mathematical morphology

Atherosclerosis causes millions of deaths and billions in expenses worldwide. Intravascular Optical Coherence Tomography (IOCT) is an intravascular imaging modality, used in coronary visualization and neo-intima post stent re-stenosis investigation. Segmentation is important for the re-obstruction quantification, improving the overall procedures. As IOCT is relatively new, few fully automatic stent segmentation works can be found in the literature. Since IOCT provides hundreds of images, non-automatic segmentation procedures may be an arduous task. Consequently, we present a fully automatic stent segmentation methodology, based on a combination of contrast stretching; wavelet decompositions as Feature Extraction; and morphological reconstruction used as post-processing so as to select and improve the previous obtained information. The evaluation was performed by segmenting 160 images from pig coronaries, containing a variety of stent disposition; hence, the outcomes were compared with their corresponding gold standards. The final results led to: True Positive (%) = 93.35±6.49, and False Positive (%) = 8.05±11.6‥ The outcome provided accurate values; in addition, it is a complete automatic approach.