A non-invasive screening technique for visualizing coronary arteries which overcomes the problems of visualizing the curved arteries by projecting the three dimensional volume of the arteries onto a two dimensional screen. Blood-filled areas such as the coronary arteries and veins, are highlighted to contrast with other nearby tissues using non-linear classification and segmentation techniques. Data is gathered as a sequence of 2D slices stored as a 3D volume. Software interpolates voxels intermediate to the slices. Wiener filtering or LMS spatial filtering can be implemented on each 2D scan to improve lateral resolution and reduce noise prior to the use of the scan data with the classification and segmentation algorithms. A traditional handheld ultrasound probe is employed to enable the technician to locate the area of interest, but a gyroscopic stabilizer is added to minimize unwanted variation on two axes of rotation.