MRI Radiomics Model Predicts High-Grade Prostate Cancer in a Two-Center Study
Published in Artificial Intelligence, Journals, Scanners and Imaging.
Published October 5, 2026. Source date: October 2, 2026.
Researchers reported a machine-learning framework combining multiparametric MRI, features from regions within and around prostate tumors, and clinical information to estimate high-grade cancer before surgery. The peer-reviewed study appeared in Frontiers in Oncology.
The retrospective analysis included 274 patients: 145 for training, 63 for internal validation and 66 from a second center for external testing. The combined model achieved areas under the ROC curve of 0.914, 0.903 and 0.797, respectively. These values describe discrimination, rather than percentages of correct diagnoses.
Features from a 5-mm region around the tumor performed best among the surrounding regions tested. The combined model was not significantly better than the habitat model alone. Lower performance in the external cohort highlights the need for broader validation; the findings do not establish that this research tool improves treatment outcomes.