Logo image
Back
MRI Deep Learning for Differentiating Glioblastoma, IDH-Wildtype from Central Nervous System Diffuse Large B-cell Lymphoma
Journal article   Open access   Peer reviewed

MRI Deep Learning for Differentiating Glioblastoma, IDH-Wildtype from Central Nervous System Diffuse Large B-cell Lymphoma

Mana Moassefi, Paul A Decker, Gian Marco Conte, Matthew L Kosel, Annette M Molinaro, Moritz A Niederschweiberer, Yalda Nikanpour, Michael W Ruff, Terry C Burns, Umar Farooq, …
Cancer research communications
05/04/2026
DOI: 10.1158/2767-9764.CRC-25-0710
PMID: 42081255
url
https://doi.org/10.1158/2767-9764.CRC-25-0710View
Published (Version of record) Open Access

Abstract

Glioblastoma, IDH-wildtype (GBM) and central nervous system diffuse large B-Cell lymphoma (CNS-DLBCL) are aggressive brain tumors with overlapping MRI features, yet distinct treatment approaches. Non-invasive tools are needed to aid in differential diagnosis. Deep learning on T1 post-contrast and T2-weighted MRI sequences were used to differentiate GBM and CNS-DLBCL. A three-stage temporal study design was utilized. Model development was performed on 146 CNS-DLBCL and 146 age, sex, and MRI year matched GBM patients diagnosed at Mayo between 1998-2019. Models were tested on independent temporal test cohorts. Initial testing included 240 independent GBM diagnosed at Mayo between 1998-2019. The prospective test cohort included 37 CNS-DLBCL and 256 GBM diagnosed at Mayo after January 1, 2020, and 36 CNS-DLBCL diagnosed at an external institution. Of the patients diagnosed at Mayo, 47% had MRIs generated from non-Mayo institutions. Two different model approaches were compared: (i) ensemble approach using AUC and cross validation for model selection, and (ii) loss approach minimizing cross entropy loss and cross validation to evaluate prediction performance. AUC on the prospective test cohort was 0.84 (95% CI: 0.78-0.90) and 0.83 (95% CI: 0.77-0.88) for the ensemble and loss approach, respectively. Stability of ensemble prediction improved with increasing number of models. Stratified AUC analysis demonstrated consistent performance across sex and age. We utilized a robust temporal study design and applied two different analytical approaches to develop a classification model. The findings confirm the feasibility of using MRI-based deep learning models to differentiate GBM from CNS-DLBCL.

Details

Metrics

1 Record Views
Logo image