Linear discriminant analysis of regional ejection fractions in the diagnosis of coronary artery disease.

We studied the application of linear discriminant analysis to the computer assisted interpretation of rest and exercise gated blood pool studies. Data were obtained in 45 patients in whom the presence or absence of coronary artery disease was determined by coronary angiography. Automated analysis of regional ejection fractions using a linear discriminant function (LDF) was compared to the fully automated calculation of global ejection fraction and to the subjective evaluation of wall motion. A LDF constructed from the data for the first 31 consecutive patients correctly classified 25/31 (81%). The change in global ejection fraction correctly classified 24/31 (77%) and wall motion analysis 25/31 (81%). In a distinct group of 14 patients the same LDF correctly classified 12/14 (86%), global ejection fraction 9/14 (64%) and wall motion analysis 10/14 (71%).