AI-Powered Phones Help Midwives Hear Fetal Heartbeats Clearly

Researchers have unveiled a technical framework that brings advanced prenatal care to rural communities using little more than an Android phone. Co-designed with Indigenous midwives in rural Guatemala, the system uses 'edge-AI'—artificial intelligence that processes data locally on the device rather than sending it to a central server—to analyse fetal Doppler signals.

The deep neural network listens to audio recordings in 3.75-second intervals, classifying them into five categories, such as 'good', 'poor', or 'radiofrequency interference'. This allows the app to provide immediate feedback, ensuring midwives capture clear heart rate data essential for detecting conditions like fetal growth restriction or hypertension. In tests, the model achieved 99.2% accuracy for identifying 'good' quality data within the Guatemalan dataset and successfully generalised to data collected in a German hospital. By securing high-quality data at the source, this scalable mobile health solution promises to enhance maternal monitoring significantly in the Global South.