Reliability and normal variations of computer analysis of Frank electrocardiogram by Smith-Hyde program (1968 version).

Abstract A preliminary assessment was made of automated electrocardiographic signal recording and telephone transmission for recording on magnetic tape and subsequent "off-line" computer analysis by the initial Smith-Hyde program (1968) of the Frank X, Y and Z leads in 334 healthy persons and 39 cardiac patients obtained from 2 hospital populations in Seattle. The Smith-Hyde analytical program provided an over-all sensitivity of 76 percent for true positives (abnormal electrocardiograms) and specificity of 75 percent for true negatives (normal electrocardiograms) for predominantly normal case material used in this study. Although 25 percent were misclassified, the ratio of correct to incorrect classifications was 3.0 for all items, and ranged from 4.7 to 61.2 for various diagnostic categories. Angular components of the quantified normal electrocardiogram showed less variation than voltage magnitudes, and bimodal distributions for initial and terminal QRS forces (ROR 3 and RER 4 ) were found only in the frontal plane when differences in sex and age were disregarded. There were more significant differences in voltage magnitudes of ventricular repolarization than of either atrial or ventricular depolarization in relation to sex and age of healthy persons.