On-line classification of arterial stenosis severity using principal component analysis applied to Doppler ultrasound signals

Abstract

Principal component analysis is a powerful method of feature extraction which can be applied to continuous-wave Doppler waveforms. A microprocessor system for the on-line calculation of the coefficients of principal components has been devised and tested in an experimental model. Doppler waveforms were obtained from positions distal to stenoses of known severity implanted in the iliac arteries of three dogs and classified into one of four groups. By reference to data from a previous series of experiments the microprocessor correctly classified 75% of stenoses. The remaining 25% were all classified as being one group more severe than they actually were.