Development and external validation of an ultrasound image-based deep learning model to estimate gestational age in the second and third trimesters of pregnancy using data from Garbh-Ini cohort: a prospective cohort study in North Indian population

Abstract

Accurate estimation of gestational age (GA) is essential to plan appropriate antenatal care. Current GA estimation models rely on fetal biometry measurements, which are susceptible to ethnic and pathological variations in fetal growth, especially in the second and third trimesters of pregnancy. In this study, we challenge the current paradigm of estimating GA using fetal biometry, by using ultrasound (US) images and deep learning models which can automatically learn image features associated with GA. We developed deep learning models for GA estimation using US images taken at 18-32 weeks of pregnancy from 2207 participants of Garbh-Ini - a hospital-based prospective cohort of pregnant women in North India. Further, we designed a novel conformal prediction (CP) algorithm to detect and reject images when there is a data distribution shift, preventing erroneous predictions. Our best model, GArbh-Ini Ultrasound image-based Gestational age Estimator (GAUGE), which was trained on US images of the fetal head (9647 images from 2207 participants), had a mean absolute error (MAE) of 2.8 days when evaluated on an internal test dataset (N = 204). GAUGE is 44% and 35% more accurate than the widely used Hadlock and INTERGROWTH-21st biometry-based GA models, respectively on the internal test dataset. For an external test dataset (N = 311), collected retrospectively from The Ultrasound Lab, New Delhi, the same model achieved a MAE of 5.9 days. In addition, we show that GAUGE relies on the finer details in the image instead of the fetal biometry and that this leads to a similar performance across small for gestational age (SGA) and appropriate for gestational age (AGA) groups. The ability of GAUGE to consider image features beyond derived biometry suggests that GAUGE offers a better choice for populations with a high prevalence of fetal growth restriction.