Computer assessment of the intrapartum cardiotocogram. I. Methods of data reduction and diagnostic procedure

To solve the problem of low predictive values in the visual assessment of the CTG, several computer programs have been designed. Only a few of those programs describe the CTG automatically and thus consistently. The aim of this study was to construct a diagnostic procedure for computer-aided automatic assessment of the CTG. A computer program quantified 17 variables of the CTG. Within a window of user-defined length, a second program calculated the number of measurements, the mean, the trend, and the coefficient of variation of each of the 17 variables, and produced a total of 56 distinct subvariables. A discriminant procedure was set up which could use any number and combination of the 56 subvariables to calculate the probability of a CTG derived from a compromised infant. The diagnostic procedure was optimized by "direct forward search" for the combination of subvariables discriminating best between normal and compromised infants. Using the last half hour of the first stage of labor, 50 CTGs entered the optimizing procedure. A combination of nine subvariables was found. In a separate paper its performance is compared with that of clinicians.