Why teaching AI about symmetry could make prenatal diagnosis more reliable

Researchers at MIT, Boston Children's Hospital and the French institute Inria have built a deep learning model, E(3)-Pose, that tracks a fetus's head during MRI so scanners can adapt as it moves, MIT News reports. The work, published October 19, 2026 by the university's news office, was presented at the 2026 European Conference on Computer Vision.

Fetal MRI, often ordered after an ultrasound flags a possible brain problem, builds a 3D picture from stacked 2D slices. When the head shifts, slices land at odd angles and leave gaps that can cause physicians to miss or misread a pathology. E(3)-Pose builds the left-right symmetry of the fetal head directly into its network architecture rather than leaving the model to infer it from data, which makes the brain's orientation easier to identify.

First author Ramya Muthukrishnan, a PhD student in Polina Golland's lab, says she crashed the scanner repeatedly while moving the method from GPU training to real hardware. She was motivated by watching a pregnant patient abandon a scan after an hour. Ellen Grant, who directs a fetal imaging center at Boston Children's, says the work will automate a slow manual process, yielding faster, more comfortable scans and better images.

The model is trained primarily on third-trimester fetuses, and Muthukrishnan hopes to extend it to younger ones, which move more and where earlier detection makes intervention easier. The researchers also suggest uses in other hard-to-image anatomy and in autonomous driving.