Left ventricular wall segmentation in ultrasound cross-sectional images

When ultrasound images are examined, identification of various tissues and organs is required. In cardiac cross sectional images, distinguishing between myocardial tissue and blood is mandatory, and should preferably be performed automatically. Here, the problem of coarse detection of cardiac myocardial boundaries in short axis images is addressed using a segmentation method. Due to the complexity of this problem, time-space algorithms are used. Moreover, processing of an ultrasound image is based upon a multiresolution presentation of the image, using the discrete wavelet transform (DWT), which results in the segmentation of the image into blood and tissue areas. Heuristic time-space algorithms for improvement of the segmentation follow this stage. Comparison of the results to those of an expert cardiologist shows good agreement.