Boundary detection in carotid ultrasound images using dynamic programming and a directional Haar-like filter

The intima–media thickness (IMT) of the carotid artery, obtained from B-mode ultrasound images, has recently been proposed as one of the most useful indices of atherosclerosis and can also be used to predict major cardiovascular events. Ultrasonic measurements of the IMT are conventionally obtained by time-consuming manual tracing of the interfaces between tissue layers. We propose a computerized method to detect the boundary of the intima–media complex using a directional Haar-like filter that can account for the slope of the boundary in an image. The directional Haar-like filter extracts a directional boundary feature as an image feature in the region of interest, which is used to compute a cost function. A cost function includes not only the directional Haar-like filtering value but also the geometric continuity that is computed for every pixel in the region of interest. The optimal boundary pixels are detected by using a dynamic programming approach that searches for the pixel that minimizes the cost function in each column of the image. We compared the performance of the proposed method with that of manual methods performed by two radiologists. The results showed that our approach produces very similar results to those based on manual tracing, and there was no statistically significant difference between the IMT measurements segmented manually and those analyzed using our method.