The Application of Classification Techniques to Biomedical Data, with Particular Reference to Ultrasonic Doppler Blood Velocity Waveforms

Abstract
The automatic classification of waveforms, derived from the body, into normal and abnormal, or into those with different degrees of abnormality, is becoming more feasible as microcomputer systems become readily available. There are many ways to achieve such classifications, but the best methods in any instance will depend on many factors. In this study, several methods have been used to classify ultrasonic Doppler blood velocity waveforms, derived from humans with peripheral arterial disease and dogs with stenoses of different severities implanted into their hind limbs. The results have been compared, and the advantages and disadvantages of the methods are discussed.