Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Wednesday, July 01, 2020 at 8:00 PM to 10:00 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Accepted for/Published in: Journal of Medical Internet Research

Date Submitted: Jun 9, 2026
Date Accepted: Oct 22, 2026

The final, peer-reviewed published version of this preprint can be found here:

Diagnostic Performance of Radiomics-Based Machine Learning in Differentiating Glioblastoma From Brain Metastasis: Systematic Review and Meta-Analysis

Hong C, Liu G, Bie Z, Guo Y, Du Y, Hu Y, Li S, Liu P

Diagnostic Performance of Radiomics-Based Machine Learning in Differentiating Glioblastoma From Brain Metastasis: Systematic Review and Meta-Analysis

J Med Internet Res 2026;28:e104060

DOI: 10.2196/104060

Diagnostic Performance of Radiomics-Based Machine Learning in Differentiating Glioblastoma from Brain Metastasis: A Systematic Review and Meta-Analysis

  • Chen Hong; 
  • Gemingtian Liu; 
  • Zhixu Bie; 
  • Yuanhang Guo; 
  • Yi Du; 
  • Yanan Hu; 
  • Shiwei Li; 
  • Pinan Liu

ABSTRACT

Background:

Glioblastoma (GBM) and brain metastasis (BM) are the most common malignant intracranial tumors in adults. Non-invasive differentiation between them remains challenging, as they frequently share similar imaging features. Although radiomics-based machine learning (ML) models show promise, systematic evidence of their performance remains limited.

Objective:

We aimed to systematically estimate the diagnostic performance of radiomics-based ML models in differentiating GBM from BM and to explore multiclass classification involving primary central nervous system lymphoma (PCNSL), assessing their potential for clinical utilization.

Methods:

We systematically searched PubMed, Embase, Web of Science, and Cochrane Library from inception to January 22, 2026, for studies reporting the diagnostic performance of radiomics-based ML models for differentiating GBM from BM, with multiclass studies involving PCNSL also retained. Methodological quality was appraised via the Radiomics Quality Score 2.0. A bivariate mixed-effects model was used for binary tasks and a random-effects model for multiclass tasks.

Results:

Forty-one studies (12,712 individuals) were included. In internal validation, summary ROC AUC was 0.92 (95% CI: 0.78–0.97), with pooled sensitivity of 87% (95% CI: 82%–91%) and specificity of 84% (95% CI: 80%–88%). Performance declined in external validation [AUC: 0.87 (95% CI: 0.67–0.95); sensitivity: 80% (95% CI: 69%–88%); specificity: 82% (95% CI: 76%–87%)]. Models outperformed less experienced experts [experience <10 years; sensitivity: 75% (95% CI: 68%–81%); specificity: 71% (95% CI: 65%–76%)] and were comparable to highly experienced experts [≥10 years; sensitivity: 85% (95% CI: 81%–89%); specificity: 82% (95% CI: 76%–87%)]. Multiclass accuracy for GBM, BM, and PCNSL was 87% (95% CI: 82%–92%), 78% (95% CI: 68%–87%), and 69% (95% CI: 57%–80%), respectively.

Conclusions:

Radiomics-based ML models demonstrate moderate accuracy in differentiating GBM from BM, outperforming less experienced specialists in pooled indirect comparisons. However, inherent methodological biases and the absence of robust external validation currently preclude clinical utilization. Multicenter, prospective validations in consecutive patient populations are required. Clinical Trial: PROSPERO CRD420261295269.


 Citation

Please cite as:

Hong C, Liu G, Bie Z, Guo Y, Du Y, Hu Y, Li S, Liu P

Diagnostic Performance of Radiomics-Based Machine Learning in Differentiating Glioblastoma From Brain Metastasis: Systematic Review and Meta-Analysis

J Med Internet Res 2026;28:e104060

DOI: 10.2196/104060

Download PDF


Request queued. Please wait while the file is being generated. It may take some time.

© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.