Intravascular ultrasound video processing for blood vessel measurement

Methods for extracting features of coronary arteries are important to assist accurate diagnosis, and the selection of regions for treatment. The authors proposed a variational model based method to detect closed boundary automatically from an intravascular ultrasound image. They then apply this method to detect lumen-intima boundary and intima-media boundary directly from a sequence of frames of a digitized intravascular ultrasound video, which was recorded in vivo at 30 frames/sec during a constant speed (0.5 mm/sec) drawback in a pig LAD artery. The authors utilize temporal and geometry continuity between the neighboring frames to speed up the boundary detection process from video frames. They then define a set of morphological parameters, such as vessel area, lumenal area, circularity, and intimal thickening, and calculate them from the sequence of the detected lumen-intima and intima-media boundaries. The authors separate these parameters into steady and transient part and use auto-correlation to extract heart beat frequency from the transient part of the vessel area function, this frequency matches very well with clinical record, the error being 1 percent