AutoHyPE: AI Detects Maternal Hypertension via Fetal Doppler Ultrasound

A new hierarchical attention network named AutoHyPE has been created by researchers to evaluate one-dimensional Doppler ultrasound recordings of fetuses for identifying maternal hypertension. This innovative system was trained on a dataset comprising 3,255 pregnant women during 8,170 antenatal visits in rural Guatemala. Utilizing prototype-based contrastive learning along with a multi-view approach, AutoHyPE effectively addresses long-tailed class distribution and biological variability. The model achieved an impressive A score, indicating its potential for continuous, non-invasive hypertension screening at the point of care. This method overcomes the shortcomings of intermittent cuff-based measurements, which do not reflect continuous physiological changes. The findings indicate that fetal cardiovascular activity may reveal indicators of maternal hypertension, paving the way for new diagnostic methods for hypertensive disorders during pregnancy, a major contributor to maternal and fetal health issues globally.