Introduction Multiple radiomics-based models have been proposed for glioma grading with different magnetic resonance imaging sequences, models, and features.
Aim of Study Given the heterogeneity and rapid expansion of radiomics for glioma grading, we aimed to better define the overall performance of these different techniques.
Methods We conducted a systematic review of the literature and a meta-analysis of studies reporting on radiomics for glioma-grade prediction. A comprehensive literature search of the databases PubMed, Ovid MEDLINE, and Ovid EMBASE was designed and conducted by an experienced librarian with input from the authors. We estimated overall sensitivity (SEN) and specificity (SPE). Event rates were pooled across studies using a random-effects meta-analysis, and the χ2 test was performed to assess the heterogeneity.
Results Overall SEN and SPE for differentiation between low-grade glioma (LGG) and high-grade glioma (HGG) were 91% and 84%, respectively. As for the discrimination task between WHO grade III and WHO grade IV, the overall SEN was 89% and the overall SPE was 81%. There is a better trend for modern non-linear classifiers while textural features are the most used and the best-performing (28.6%).
Conclusion The current diagnostic performance of radiomics for glioma grading is higher for the LGGs vs. HGGs discrimination task than the WHO grade III vs. IV task, both in terms of SEN and SPE. In the forthcoming years, we expect even more precise models, especially for the LGGs vs. HGGs categorization.
Disclosure of Interest Nothing to disclose
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.