UChicago receives $3.1 million NIH grant to build AI-generated pediatric imaging database
The planned public platform would give researchers and surgeons synthetic reference images while addressing privacy barriers that limit access to real scans.
Loading page…
The planned public platform would give researchers and surgeons synthetic reference images while addressing privacy barriers that limit access to real scans.
Listen to this story
The University of Chicago plans to turn curated pediatric scans into synthetic reference datasets, then make them available through a public graphical platform; the project is funded by a $3.1 million NIH grant and led by neurological surgeon Antonio Porras. The planned collections cover healthy skull development and craniosynostosis before and after surgery. The resource is not yet available, and its proposed uses—supporting surgical comparisons and development research without sharing real patient scans—are goals, not demonstrated clinical outcomes.
The project will build the healthy-development collection first, followed by a craniosynostosis repository with pre- and post-surgery images.
The planned interface is meant to let surgeons compare a patient with reference data without needing specialized data skills.
Synthetic images must be both realistic and scientifically accurate for other institutions to rely on them in research.
Children’s head scans can help researchers understand how skulls grow, but privacy concerns can prevent hospitals from sharing them. The University of Chicago has received a $3.1 million National Institutes of Health grant to develop AI-generated alternatives, with a planned public platform intended to support cranial-development research and help surgeons evaluate patients.
The university described the project in its November 12 announcement. Researchers will use a curated collection of clinical pediatric images to develop generative AI tools that produce realistic synthetic images across ages and stages of skull development. The intended output is a reference resource other institutions can use, rather than a collection of real patient scans passed between hospitals.
Antonio Porras, an associate professor of neurological surgery at UChicago, will lead the work. Partners include Children’s National Hospital in Washington, D.C.; Children’s Hospital Los Angeles and the University of Southern California; and Children’s Hospital Colorado and the University of Colorado Anschutz Medical Campus. Together, they will develop datasets representing both healthy children and children with a condition called craniosynostosis.
Craniosynostosis occurs when the seams between bones in an infant’s skull fuse too early, before the brain has fully formed. Surgeons can reopen the prematurely fused bones to give the brain room to develop. Planning those procedures requires assessing the skull’s current shape and structure, then projecting how it will develop after an operation.
That projection has to account for rapid growth. A child’s skull can change in the few weeks between an initial scan and surgery. The team wants surgeons to compare patients with reference images showing expected development at the same stage. UChicago says that could improve preparation without requiring additional scans that would expose children to more radiation.
Synthetic images are the proposed way around the sharing barrier. Porras says pediatric-focused AI methods are beginning to let his team generate datasets without compromising privacy. He describes the scientific task as learning to make those datasets both realistic and scientifically accurate, so other researchers can use them for their own studies of craniofacial development.
Not everyone works at a large, high-volume academic medical center or has access to experienced surgeons who have done thousands of these procedures.
Rohaid Ali, assistant professor of neurological surgery at UChicago
Porras laid out a sequence for the project, with the image collections coming before the public interface. The work remains a development effort, not a released database or an available surgical-planning service. Its announced stages are:
The before-and-after collection has a different purpose from the healthy-child reference set. Porras says the team wants to quantify children’s skull abnormalities at diagnosis and forecast their development. It also aims to let surgeons compare outcomes from different surgical approaches, helping them assess which procedure may be best suited to an individual patient. Those are intended uses of the project, not reported clinical results.
The final interface is intended to make that resource usable without specialized technical data skills. For hospitals without those skills on staff, the planned access is not simply permission to download images: surgeons would have a graphical tool for comparing the shared reference data with the patients they are treating.
Loading discussion...
Join the conversation
Explain what would build or undermine your confidence.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.