Researchers trained and validated a deep learning model that can detect subtle changes across post-treatment brain scans and forecast glioma recurrence with up to 89 percent accuracy.