Artificial Intelligence Tool Can Differentiate Glioblastomas From CNS Lymphoma

THURSDAY, Oct. 2, 2025 (HealthDay News) — A Pathology Image Characterization Tool with Uncertainty-aware Rapid Evaluations (PICTURE) system can differentiate glioblastomas from primary central nervous system (CNS) lymphoma, according to a study published online Sept. 29 in Nature Communications.

Junhan Zhao, from Harvard Medical School in Boston, and colleagues established the PICTURE system using 2,141 pathology slides collected worldwide to address the challenge of distinguishing glioblastoma and primary CNS lymphoma. To account for the uncertainties in its predictions and training set labels, PICTURE employs Bayesian inference, deep ensemble, and normalizing flow.

The researchers found that glioblastoma and primary CNS lymphoma were accurately diagnosed with PICTURE, with an area under the receiver operating characteristic curve (AUROC) of 0.989; the results were validated in five independent cohorts, with AUROCs ranging from 0.924 to 0.996. Samples belonging to 67 types of rare CNS cancers were also identified by PICTURE, which were neither gliomas nor lymphomas.

“Our model can minimize errors in diagnosis by distinguishing between tumors with overlapping features and help clinicians determine the best course of treatment based on a tumor”s true identity,” senior author Kun-Hsing Yu, from Cedars-Sinai Medical Center in Los Angeles, said in a statement.

Abstract/Full Text