A digital computer program is described for automatic identification of the clinically important points in the externally recorded brachial pulse wave. The logic of recognition is designed to locate these points in any type of brachial pulse wave. Intrinsic to the program is the identification of possible errors of recognition at each step. In 332 brachial pulse wave records comparisons between statistics calculated from the computer identified points with corresponding measurements calculated from points identified by an experienced physician showed correlation coefficients varying between .94 and .99 indicating good agreement.
PATTERN RECOGNlTION IN THE BRACHIAL PULSE WAVE
Externally recorded arterial pulse waves provide a relatively simple, nonbloody method for assessing certain cardiovascular abnormalities. They have been used for many years in the diagnosis of aortic valvular disease and for the estimation of pulse wave velocity. Recordings of arterial pulse waves have been employed in assessing the functional status of the heart 1 and in the estimation of arterial changes associated with aging and vascular disease.".:'
Computer techniques have been utilized in pulse wave analysis but the methods involved considerable manual processing in addition to that done by the computer." By developing a pattern recognition program the processing of pulse waves becomes almost completely automated.
This report describes a program for automatic identification of the clinically important points in the brachial pulse wave cycle and presents the results of comparisons between statistics calculated from these points with corresponding measurements calculated from points visually identified by an experienced physician, METHODS Brachial pulse waves were recorded on FM magnetic tape from a sample of 350 male subjects ranging in age from 15 to 88 years. The sample was comprised