Machine Learning-Based Prediction of Glioma IDH Gene Mutation Status Using Physio-Metabolic MRI of Oxygen Metabolism and Neovascularization - A Bicenter Study

Aug 6, 2026

The mutational status of the isocitrate dehydrogenase (IDH) gene is critical in glioma management because it influences energy metabolism pathways. Physio-metabolic MRI provides a non-invasive way to analyze oxygen metabolism and tissue hypoxia, linked to neovascularization and microvascular architecture, but complex neuroimaging needs computational support. Traditional machine learning and simple deep learning models were trained on radiomic features from clinical MRI and physio-metabolic MRI. Across 215 patients from two centers and using independent internal and external testing, physio-metabolic models performed best internally, while clinical MRI-trained models performed better externally. Results highlight the need for standardized protocols and robust independent validation for clinical MRI-driven AI.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Overview

Clinical significance of IDH mutation status

Imaging and computational approach

Study design

Patient cohort and MRI protocols

Training and testing strategy (internal vs external)

Modeling methods

Traditional machine learning

Deep learning models

Feature sources (clinical MRI vs physio-metabolic MRI)

Results

Internal testing performance metrics

External testing performance comparison

Discussion and implications

Site-dependent acquisition effects

Standardization and reproducibility needs

cancers Article Machine Learning-Based Prediction of Glioma IDH Gene Mutation Status Using Physio-Metabolic MRI of Oxygen Metabolism and Neovascularization (A Bicenter Study) Andreas Stadlbauer 1,2,3, *, Katarina Nikolic 1,4, Stefan Oberndorfer 1,4, Franz Marhold 1,5, Thomas M. Kinfe 3,6, Anke Meyer-Bäse 7, Diana Alina Bistrian 8, Oliver Schnell 3 and Arnd Doerﬂer 9 Citation: Stadlbauer, A.; Nikolic, K.; Oberndorfer, S.; Marhold, F.; Kinfe, T.M.; Meyer-Bäse, A.; Bistrian, D.A.; Schnell, O.; Doerﬂer, A. Machine Learning-Based Prediction of Glioma IDH Gene Mutation Status Using Physio-Metabolic MRI of Oxygen Metabolism and Neovascularization (A Bicenter Study) . Cancers 2024, 16, 1102. [https://doi.org/10.3390/](https://doi.org/10.3390/)[ ](https://doi.org/10.3390/)cancers16061102 Academic Editor: David Wong Received: 29 January 2024 Revised: 21 February 2024 Accepted: 23 February 2024 Published: 8 March 2024 Copyright: © 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . 1 Karl Landsteiner University of Health Sciences, 3500 Krems, Austria; katarina.nikolic@stpoelten.lknoe.at (K.N.); [stefan.oberndorfer@stpoelten.lknoe.at](stefan.oberndorfer@stpoelten.lknoe.at) (S.O.); [franz.marhold@stpoelten.lknoe.at](franz.marhold@stpoelten.lknoe.at) (F.M.) 2 Institute of Medical Radiology, Diagnostics, Intervention, University Hospital St. Pölten, 3100 St. Pölten, Austria 3 Department of Neurosurgery, Universitätsklinikum Erlangen, Friedrich-Alexander University (FAU) Erlangen-Nürnberg, 91054 Erlangen, Germany; [thomasmehari.kinfe@uk-erlangen.de](thomasmehari.kinfe@uk-erlangen.de) (T.M.K.); [oliver.schnell@uk-erlangen.de](oliver.schnell@uk-erlangen.de) (O.S.) 4 Division of Neurology, University Hospital St. Pölten, 3100 St. Pölten, Austria 5 Division of Neurosurgery, University Hospital St. Pölten, 3100 St. Pölten, Austria 6 Division of Functional Neurosurgery and Stereotaxy, Friedrich-Alexander University (FAU) Erlangen-Nürnberg, 91054 Erlangen, Germany 7 Department of Scientiﬁc Computing, Florida State University, 400 Dirac Science Library Tallahassee, Tallahassee, FL 32306-4120, USA; [ameyerbaese@fsu.edu](ameyerbaese@fsu.edu) 8 Department of Electrical Engineering and Industrial Informatics, Politehnica University of Timisoara, 300006 Timis, oara, Romania; diana.bistrian@ﬁ[h.upt.ro](h.upt.ro) 9 Department of Neuroradiology, Universitätsklinikum Erlangen, Friedrich-Alexander University (FAU) Erlangen-Nürnberg, 91054 Erlangen, Germany; arnd.doerﬂ[er@uk-erlangen.de](er@uk-erlangen.de) * Correspondence: [andi@nmr.at](andi@nmr.at); Tel.: +43-2742-9004-14185; Fax: +43-2742-9004-49300 Simple Summary: Early characterization of the isocitrate dehydrogenase (IDHIDH) gene mutation status of glioma is crucial for personalized decision making and prognosis in clinical neurooncological treatment. Based on the known differences in energy metabolism between IDHIDH-mutated and IDHIDH-wildtype gliomas, we assessed physio-metabolic magnetic resonance imaging-based measures along with machine learning for potential reliable presurgical characterization of IDHIDH gene status. Traditional machine learning algorithms and simple deep learning models trained in analyzing physio-metabolic parameters demonstrated the best performance in classifying the IDHIDH gene status of gliomas in independent internal testing. In contrast, external testing revealed that traditional machine learning models trained on clinical MRI data had higher accuracy compared to physio-metabolic algorithms, reﬂecting differences in data acquisition methodology between the two sites. Our results outline the necessity of independent internal and extern

FAQs

Why is predicting IDH gene mutation status important in glioma?

IDH mutation status guides personalized clinical decisions and prognosis because it affects energy metabolism pathways relevant to glioma treatment.

What imaging modalities and features were used for the models?

Models used radiomic features derived from clinical MRI (cMRI) and physio-metabolic MRI, which targets oxygen metabolism and hypoxia-related signals and associated vascular architecture.

How did model performance differ between internal and external testing?

Physio-metabolic MRI-trained algorithms showed the best classification performance in independent internal testing, whereas external testing favored traditional machine learning models trained on clinical MRI data, indicating differences in data acquisition between sites.

Aug 6, 2026

Upload by Lucas Martin