Hierarchical Voting-Based Feature Selection and Ensemble Learning Model Scheme for Glioma Grading with Clinical and...

Area of Science:

• Neuro-oncology
• Computational Biology
• Machine Learning

Background:

• Glioma grading is crucial for optimizing patient treatment and survival rates.
• Molecular markers are increasingly vital for accurate tumor classification.
• Current grading methods can be improved with advanced computational approaches.

Purpose of the Study:

• To develop a novel hierarchical voting-based methodology for glioma grading.
• To enhance feature selection and machine learning model performance using clinical and molecular predictors.
• To improve the accuracy and efficiency of glioma classification for clinical decision-making.

Main Methods:

• Utilized publicly available TCGA and CGGA datasets for glioma grading.
• Employed a novel hierarchical voting-based ensemble feature selection method.
• Compared sixteen combination sets of four dimensionality reduction methods and five supervised models.
• Evaluated performance against the LASSO feature selection method.

Main Results:

• The proposed hierarchical voting method achieved 87.606% accuracy on the TCGA dataset.
• The method reached 79.668% accuracy on the CGGA dataset.
• Outperformed the isolated LASSO feature selection method in accuracy.

Conclusions:

• The novel hierarchical voting-based methodology significantly improves glioma grading accuracy.
• This approach enhances the utility of molecular and clinical data for tumor classification.
• The findings support the potential of this method for streamlined clinical decision-making in neuro-oncology.