MRI model tracks a moving fetus's head to avoid gaps in brain scans, team says

A team at MIT, Boston Children's Hospital and Inria built a model that tracks a moving fetal head during an MRI scan. It was presented as a paper at the 2026 European Conference on Computer Vision.

Share
MRI model tracks a moving fetus's head to avoid gaps in brain scans, team says
Illustration: AI-generated

A moving fetus can leave gaps in an MRI scan of its brain. Researchers at MIT in the United States, Boston Children's Hospital and Inria have built a model called E(3)-Pose that tracks the head as it moves. The aim is a high-resolution image a radiologist can examine, according to MIT News.

In a routine prenatal ultrasound, the person doing the scan may flag possible brain abnormalities for an additional MRI. But imperfect imaging can lead to missed or incorrect diagnoses.

An MRI takes a series of flat 2D images, called slices, and stacks them into one 3D image of the head. If the head shifts, each slice may be taken from a different angle, which leaves gaps in the final scan.

"One of the problems with having the baby move is that the slices are all over the place," said first author Ramya Muthukrishnan, an MIT PhD student in the lab of Professor Polina Golland. She said coverage gaps hide critical regions of the brain needed to detect abnormal development.

Muthukrishnan's motivation came while shadowing a radiologist at Boston Children's, when she saw a pregnant mother quit a scan halfway after an hour in the MRI, leaving the radiologist without a full, high-quality image.

E(3)-Pose is a deep learning model, a program trained on examples. The team built the left-right symmetry of the fetal head into its design, which MIT News says makes it easier for the model to identify the brain's orientation. MIT News says the model adjusts scans on the fly to track fetal movement.

The team ran the model on a real MRI scanner. Co-author Borjan Gagoski, of Harvard Medical School and Boston Children's, developed a scanning sequence that acquires extra scans to help train the model. Co-author Aryn Lee, a Boston Children's research assistant, recruited the pregnant volunteers and helped acquire the scans.

The paper, "Equivariant symmetry-aware head pose estimation for fetal MRI", was presented at the 2026 European Conference on Computer Vision. MIT News published its report on 19 October 2026.

Senior author Ellen Grant, who directs the Fetal Neonatal Neuroimaging and Developmental Science Center at Boston Children's, said the work will automate fetal brain MRI, with faster, more comfortable scans and higher-quality images. MIT News describes it as a possibility that the model could spare families repeat hospital visits.

The model is mainly trained on third-trimester fetuses. Muthukrishnan hopes to improve it to help image second-trimester or younger fetuses. Of those, she said: "They’re more important to focus on because the earlier you detect brain abnormalities, the easier it is to intervene."

Sources