Motivation: Fetal MRI, an important complementary modality for antenatal diagnosis, suffers from a lack of interactivity and a need for offline processing to achieve volumetric information. However, only information available during the scan can trigger further scans to provide a faster and more individualized diagnosis. Goal(s): Real-time automatic weight estimation and volumetry on the MR scanner during the time of acquisition. Approach: Deep learning networks have been combined with real-time scanner interfacing on high-performance computers to enable real-time volumetric measurements of the fetal body, head, placenta and amniotic fluid in bSSFP images. Results: Successful prospective real-time results were acquired, highlighting the achieved novel capacity. Impact: Real-time fetal weight estimation and volumetry using AI on the scanner enables faster and more individual fetal MRI - hence paving the way for enhanced antenatal diagnosis.
Silva et al. (Tue,) studied this question.