Pre-Operative Mri-Based Radiomics for Predicting Telomerase Reverse Transcriptase Promoter Mutation Status in Glioma Patients: A Systematic Review and Meta-Analysis Publisher Pubmed



Foodeh T ; Korani M A ; Teymourzadeh M ; Hassanpour Fard Khor M ; Abdollahi Z ; Hatami B ; Payandeh A ; Shojaeshafiei F ; Seyed Ebrahimi S M ; Fayazbakhsh F ; Delshad M ; Khodadadi S ; Karimi S ; Mortezaei A Show All Authors
Source: Neurosurgical Review Published:2026

Abstract

TERT promoter (TERTp) mutations shape glioma prognosis and therapy, yet tissue testing can be limited by sampling error and surgical inaccessibility. MRI-based radiomics offers a non-invasive alternative. This study aimed to quantify the diagnostic accuracy of pre-operative MRI radiomics for predicting TERTp status and compare radiomics-only, clinical-only, and combined models.We conducted a PRISMA-DTA–conformant, PROSPERO-registered systematic review and meta-analysis. PubMed, Embase, Web of Science, and Scopus were searched to 13 October 2025. Eligible studies evaluated MRI-derived radiomics models and reported accuracy on non-training data against a molecular reference standard. Risk of bias was appraised with QUADAS-AI. Bivariate random-effects models pooled sensitivity, specificity, and AUC, prioritizing external test performance when available. Fourteen retrospective studies including 2,863 patients were eligible for systematic review; 13 studies were included in the quantitative meta-analysis. MRI-only radiomics models demonstrated pooled sensitivity of 0.76 (95% CI, 0.66–0.84), specificity of 0.70 (95% CI, 0.63–0.77), and AUC of 0.79 (95% CI, 0.75–0.82), indicating moderate discriminative performance with substantial heterogeneity. Deeks’ funnel plot asymmetry test was not significant (p = 0.78). Clinical-only models yielded pooled sensitivity of 0.73 (95% CI, 0.61–0.82), specificity of 0.57 (95% CI, 0.34–0.77), and AUC of 0.73 (95% CI, 0.69–0.77). Combined radiomics-clinical models showed numerically higher pooled performance, with sensitivity of 0.78 (95% CI, 0.70–0.85), specificity of 0.76 (95% CI, 0.67–0.84), and AUC of 0.82 (95% CI, 0.79–0.85), although this finding should be interpreted descriptively rather than as definitive evidence of superiority. Subgroup analyses suggested that classifier type, validation strategy, and feature-extraction software may contribute to performance variability. Sensitivity analysis showed that the overall findings remained broadly stable after excluding the influential study. Pre-operative MRI-based radiomics shows moderate accuracy for predicting TERTp mutation status in glioma. Combined radiomics-clinical models achieved numerically higher performance, but current evidence remains limited by retrospective designs, internal validation, and methodological heterogeneity. These models should be considered adjunctive rather than replacement tools, and prospective multicenter external validation with standardized workflows is required before clinical implementation. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2026.