Deep Learning Approach for Glioma Grade Classification with 3D Wavelet Radiomics

Abstract

Gliomas are classified as high-grade glioma and low-grade glioma based on the extent of spread. The tumor grade provides essential information about the tumor's aggressiveness and malignancy, assisting physicians in prescribing the appropriate dosage of radiation and chemotherapy. Histopathological analysis of tumor tissue samples, usually obtained through biopsy, is commonly required for brain tumor grading. However, these tissue samples may not completely represent the tumor's heterogeneity, leading to potential sampling bias. To overcome this limitation and avoid the negative impact of biopsy, it becomes crucial to assess the tumor grade directly from MRI scans. Hence in this study, we propose a method that leverages a deep neural network classifier on optimally selected 3D wavelet radiomic features, extracted automatically from multisequence 3D MRI to predict the tumor grade. The proposed method for classifying high-grade glioma and low-grade glioma is evaluated on BraTS 2019 3D MRI dataset using metrics such as F1-score, precision, recall, and accuracy. The proposed method outperforms the conventional machine learning algorithms and also outperforms the state-of-the-art tumor grade classification models.