Early Prediction of Weight at Birth Using Support Vector Regression

Abstract

Birth weight (BW) is an important indicator of neonatal well–being associated with multiple adverse conditions and thus its early estimation may be crucial for timely treatment. In this paper we propose a BW estimation strategy based on Support Vector Regression of a set of multimodal maternal–fetal features obtained in pregnancy’s first trimester. The obtained results show an average difference of 250 g between the estimated and the real BW, with percentage errors below 3% in all cases. These results contrast with other reported studies, that estimate BW very close to delivery.