Detection of luminal contour using fuzzy clustering and mathematical morphology in intravascular ultrasound images.

An innovative application of fuzzy clustering and mathematical morphology for the problem of luminal contour detection in intravascular ultrasound images is presented. Median and standard deviation are used as features for segmentation process. Comparison was made with gold standard segmented images obtained from the average of images segmented by experienced medical doctors. Tests were carried out with 20 in vivo coronary images obtained from different patients. High correlation coefficients were found between lumen regions manually and automatically defined when area, mean gray level, and standard deviation of the lumen regions were compared.