Evaluation of segmentation algorithms for intravascular ultrasound images

A large number of paradigms to segment 30 MHz intravascular ultrasound (IVUS) subtraction images is evaluated. A fully automated active contour model, initialized by thresholding and preceded by filtering with a large scale median filter is a good alternative for manual delineation of both pre-dilation and post-dilation images. Results obtained with this automated method are even better than manual segmentation, which is a major step forwards in IVUS image analysis.