Objective feature extraction applied to the diagnosis of carotid artery disease using a Doppler ultrasound technique

Using the maximum frequency envelope obtained from the sonogram of the Doppler shifted frequencies a mathematical feature extraction technique has been used to provide an objective identification of the significant waveform features. The technique described is that of principal component factor analysis. The data used were taken from an eighteen month programme in which techniques based on the use of directional Doppler ultrasound were evaluated in comparison with direct percutaneous carotid angiography and arch aortography. A tracing of the vessel outlined based on the anterior-posterior view of the angiogram was used to classify each vessel segment studied. Two groups only are considered in this initial study, a normal group (25 vessel segments) and a stenosis group (18 vessel segments). The principal component analysis technique was shown to provide a superior classification of the vessel segments when compared with the more familiar A/B ratio. Nevertheless the principal diagnostic feature is shown to be the A/B ratio.