Radiomics strategy for glioma grading using texture features from multiparametric MRI

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Accurate glioma grading is crucial for patient management and molecular stratification.
• Current grading methods can be improved for better clinical outcomes.

Purpose of the Study:

• To evaluate the superiority of radiomics features from multiparametric MRI for glioma grading.
• To assess the grading potential of different MRI sequences and parametric maps.

Main Methods:

• A retrospective study involving 153 patients with Grades II, III, and IV gliomas.
• Radiomics features were extracted from multiparametric MRI (T1, T2, diffusion, ASL) using support vector machine-based recursive feature elimination.
• Support vector machine (SVM) classifiers were developed to classify low-grade glioma (LGG) vs. high-grade glioma (HGG) and Grade III vs. IV gliomas.

Main Results:

• SVM models using optimal radiomics features achieved high accuracy and AUC for glioma grading (96.8%/0.987 for LGG vs. HGG; 98.1%/0.992 for Grade III vs. IV).
• Texture features demonstrated greater effectiveness than histogram parameters for glioma grading.
• Multiparametric MRI combined with radiomics outperformed single-sequence MRI and histogram analysis.

Conclusions:

• Radiomics features derived from multiparametric MRI provide a highly effective and noninvasive method for glioma grading.
• This radiomic strategy can significantly enhance clinical decision-making for patients with gliomas of varying grades.