Active Inference for Closed-Loop Transmit Beamsteering in Fetal Doppler Ultrasound

Abstract

This work introduces a closed-loop transmit beamsteering system for fetal Doppler ultrasound, driven by an active inference agent. The agent actively reduces uncertainty in fetal heart localization - estimated from power Doppler data via sequential Monte Carlo methods - by adaptively steering the ultrasound beam. This uncertainty-aware approach maintains high-quality Doppler signals and enables robust heart rate tracking, even under challenging, low signal-to-noise conditions.