An MIT report published October 19, 2026 described a deep learning model, E(3)-Pose, developed by researchers at MIT, Boston Children's Hospital and Inria to help track fetal head motion during MRI. The work was presented at the 2026 European Conference on Computer Vision, according to the report. In fetal MRI, stacked 2D image slices form a 3D picture of the head; if the fetus moves, slices can be taken at different angles, creating coverage gaps that may lead to missed or incorrect diagnoses.

E(3)-Pose embeds left-right symmetry of the fetal head into its architecture rather than relying only on data-derived physics. First author Ramya Muthukrishnan, an MIT PhD student, said the model is primarily trained on third-trimester fetuses, and she hopes to improve it for second-trimester and younger fetuses, which move more because of uterine space.

Ellen Grant of Boston Children's said the work could automate fetal brain MRI, replacing a slow manual process with faster, more comfortable scans. Polina Golland of MIT described the approach as combining geometry concepts with deep learning. The report also mentioned possible uses beyond fetal imaging, such as other difficult anatomy and autonomous driving. Support came from MIT Jameel Clinic, the MIT Takeda Program, the Chou Family Transformative Research Fund, NIBIB and NICHD.