ORIGINAL RESEARCH article
Front. Oncol.
Sec. Neuro-Oncology and Neurosurgical Oncology
This article is part of the Research TopicImaging to Guide Treatment in Brain TumorsView all 8 articles
Explainable, Modality-Adaptive Radiomics for MGMT Methylation Prediction in High-Grade Glioma: A Decision-Curve Analysis Study
Provisionally accepted
- 1Stanford University Department of Radiology, Stanford, United States
- 2Ethniko Metsobio Polytechneio, Zografou, Greece
- 3The Cyprus Institute of Neurology and Genetics, Nicosia, Cyprus
- 4Panepistemio Patron Tmema Iatrikes, Patras, Greece
- 5University of Miami Miller School of Medicine, Miami, United States
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Objective To develop and interpret a modality-adaptive, calibrated radiomics model for non-invasive prediction of MGMT promoter methylation in high-grade gliomas (HGG) using multi-center MRI. Methods Pre-operative MRI data from UCSF-PDGM and UPENN-GBM were processed using masks focused on intratumoral and peritumoral regions. The model was trained on conventional (T1, T2, FLAIR) and advanced (DWI/ADC, ASL) MRI sequences. It employed a novel method that automatically ignores advanced modalities if they're unavailable. Following feature ranking and redundancy reduction, six classifiers were optimized and Platt-calibrated. Their performance and clinical relevance were evaluated via Decision Curve Analysis (DCA), with SHAP explaining the contribution of individual features. Results The top LightGBM-trained model using the 500 most important features achieved an AUC of 0.67, a recall of 0.90, and an accuracy of 0.72 on the held-out test set. It demonstrated excellent accuracy in avoiding missing methylated cases, while calibration improved overall clinical benefits. Conventional and advanced MRI modalities have a balanced approach to radiomic feature importance. Texture and intensity descriptors showed the most significant influence, whereas the low FLAIR wavelet(intensity) was associated with unmethylated tumors. Conclusion This explainable, modality-adaptive radiomics model identified biologically consistent MGMT-related imaging patterns and showed decision-analytic value. Future steps include external validation and integration with clinical and molecular biomarker data.
Keywords: Decision curve analysis, Explainable AI, HGG = high-grade glioma, MGMT = [6]-methylguanine-DNA methyltransferase, Radiomics
Received: 23 Oct 2025; Accepted: 15 Dec 2025.
Copyright: © 2025 Christodoulou, Vamvouras, Pitsillos, Solomou and Georgiou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Michalis F Georgiou
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.