Multicenter DSC-MRI-Based Radiomics Predict IDH Mutation in Gliomas

Area of Science:

• Neuro-oncology
• Radiomics
• Medical Imaging

Background:

• Gliomas require accurate IDH mutation status prediction for treatment.
• Current methods lack robust DSC-MRI radiomics for IDH mutation prediction.
• Multicenter studies are crucial for validating radiomics models.

Purpose of the Study:

• To develop and validate a DSC-MRI radiomics model for predicting IDH mutation status in gliomas.
• To assess the impact of image standardization on radiomics model performance.
• To explore model explainability and identify radiomic features correlated with IDH status.

Main Methods:

• A multicenter study with exploratory and validation cohorts.
• Development of radiomics models using DSC-MRI data.
• Application of image standardization techniques prior to radiomics analysis.
• Utilized LIME and SHAP for model explainability.

Main Results:

• Initial radiomics model achieved moderate prediction accuracy for IDH mutation status.
• Image standardization significantly improved prediction accuracy (0.544 to 0.706).
• Explainability methods revealed correlations between heterogeneity, texture complexity, and IDH-wildtype status.

Conclusions:

• DSC-MRI radiomics shows promise for predicting IDH mutation status in gliomas.
• Image standardization enhances the predictive performance of radiomics models.
• Radiogenomic patterns offer potential for clinical translation in neuro-oncology.