MIT model uses symmetry to track fetal movement during brain MRI scans

Researchers at MIT, Boston Children's Hospital and the French research institute Inria have developed a deep learning model called E(3)-Pose that tracks fetal movement during MRI scans of the fetal brain. When a routine prenatal ultrasound shows a possible brain abnormality, families are often referred for an MRI. Because a fetus can move during the scan, the images can come out imperfect, which can lead to missed or incorrect diagnoses and repeat appointments. E(3)-Pose adjusts scans on the fly to follow the fetus's position. Its name refers to the mathematical symmetries of three-dimensional space, and the researchers built knowledge of those symmetries into the model so it can handle a fetus that rotates or shifts. The team says this could help clinicians diagnose developmental anomalies earlier and more accurately, starting with the brain, one of the first organs to develop and one that keeps developing after birth. The work comes from MIT's Abdul Latif Jameel Clinic for Machine Learning in Health. It shows how building known physical structure into a model can make medical imaging more reliable. The announcement describes potential clinical benefits and does not report results from use in patient care.