Related Experiment Videos
Radiomics and deep learning models for predicting glioma p53 status: A diagnostic accuracy systematic review and
Amir Mahmoud Ahmadzadeh1, Mohammad Amin Ashoobi2, Nima Broomand Lomer3
1Department of Radiology, School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
Purpose:
To systematically investigate the diagnostic performance of magnetic resonance imaging (MRI)-based radiomics and deep learning (DL) models for predicting p53 status in glioma and to generate pooled estimates for radiomics-based models.
Methods:
A systematic search was conducted in PubMed, Scopus, Embase, and Web of Science. Methodological quality was appraised using the QUADAS-2 and METRICS checklists. Pooled sensitivity, specificity, positive and negative likelihood ratios, diagnostic odds ratio, and area under the receiver operating characteristic curve (AUROC) were calculated using a bivariate random-effects model. Subgroup and sensitivity analyses were undertaken to explore heterogeneity and robustness. Publication bias was assessed with Deeks' funnel plot asymmetry test.
Results:
Seventeen studies were included in the systematic review. Of these, six studies were incorporated into the meta-analysis of exclusive radiomics models and five in the meta-analysis of combined radiomics approaches. Exclusive radiomics models yielded a pooled AUROC of 0.72, whereas combined models achieved a pooled AUROC of 0.89. Studies employing manual segmentation reported a significantly greater pooled specificity compared with those applying semi-automated segmentation methods.
Conclusion:
Radiomics features showed modest performance in predicting glioma p53 status when used alone. Incorporating complementary non-radiomic features could enhance model performance. Such approaches may support non-invasive patient stratification, aid clinical decision-making and prognostic assessment, and potentially decrease reliance on high-risk biopsy procedures.
