Real-time automatic field-of-view adjustment in fetal MR scans using Gadgetron

MRI provides an ideal tool for characterising fetal brain development and growth. It is, however, a relatively slow imaging technique and therefore extremely susceptible to subject motion. To address this challenge, we are developing an intrinsically motion-robust deep-learning-based fetal MRI method to achieve real-time fetal head tracking and update the acquisition geometry prospectively. Our method uses Gadgetron for real-time reconstruction of the scans and a 3D UNet for fetal head position location and motion estimation. Real-time tracking and correction for functional fetal MRI was demonstrated both in controlled phantom setups and in fetal MRI cases.