AI-Enhanced Brain Tumor Imaging Improves Diagnostic Confidence

UCSF researchers developed an Evidential Deep Learning AI model that provides quantifiable uncertainty estimates for brain tumor MRI segmentation, specifically for meningiomas.

The model was trained on 1,655 MRIs from 788 patients and achieved high accuracy in identifying ambiguous tumor boundaries that traditional 2D metrics often fail to capture.

The study, published in npj Digital Medicine, demonstrates that providing uncertainty maps can increase clinician trust in AI-assisted diagnostics for longitudinal tumor monitoring.