[Effect of a mathematical model and number of measurements on accuracy in fetal weight estimation using ultrasound].
The problem of foetal weight estimation has not been satisfactorily solved to date. We studied quantitatively the influence of the mathematical approach and the number of distance measurements of different parts of the foetal body, on the accuracy of prospective calculation of the actual foetal weight. Of 230 foetuses of known gestational age we measured via ultrasound 13 lengths of head, rump and extremities during the last three days before delivery. 22 types of formulas were analysed. To decide on the best mathematical way of calculation we recalculated all coefficients in all formulas for each study group. In this way each formula can yield best possible results for the presented group. Three formulas yield very good results--a logarithmic type using 7 measured distances, and 2 linear types of formulas using 10 and 11 measured distances. The standard deviations of the differences (calculated weight minus actual weight) range from 216 g to 219 g. The frequency of differences greater than +/- 15% of the calculated weight ranges between 3.0% and 4.3% of the cases. Miscalculations of more than 700 g (absolutely) have not been noted. In addition, five other formulas (logarithmic and linear types with 5 to 8 measurements) have yielded good results. Using 5 to 8 measurements, the logarithmic types of formulas are better than the linear ones. The mathematical types of formulas cannot achieve the improvements resulting from a greater number of measurements.