Computer based classification of carotid arterial disease: a prospective assessment

A minicomputer based pattern recognition method has been used to prospectively classify the category of disease involvement of 105 carotid arteries. The system utilized spectral patterns obtained from a combined B-mode/pulsed Doppler unit. All decisions are based upon comparison of an unknown, averaged waveform with a series of vessels with known severity of disease. The variability in the computer decision as compared to arteriography is discussed.