AI tool improves prediction of pediatric brain tumor relapse

A new study published in The New England Journal of Medicine AI reveals that an AI model using “temporal learning” significantly improves predictions of relapse in children with brain tumors called gliomas.

Developed by researchers from Mass General Brigham, Boston Children’s Hospital, and Dana-Farber/Boston Children’s Cancer Center, the tool analyzes multiple MRI scans taken over time, rather than relying on a single image.

The study involved nearly 4,000 MRIs from 715 pediatric patients. By sequencing scans chronologically, the AI learned to detect subtle changes that may indicate recurrence. The model predicted tumor relapse with 75–89% accuracy, compared to roughly 50% using traditional methods.

Researchers say this approach may eventually reduce the burden of frequent imaging or enable early intervention for high-risk patients. Clinical trials are planned to validate its impact on patient care.