Glioma grading based on gentle-adaboost algorithm and radiomics

Glioma is one of the most important brain diseases in malignancy, mortality and morbidity in adult cancer. Accurate diagnosis of glioma is significant for treatment regimen selection. In this article, the Gentle-Adaboost algorithm and radiomics method were used to grade the gliomas. In our experiments, we firstly extracted a large number of image features. Then we adapted minimal redundancy maximal relevance criterion (mRMR) method to reduce the dimensionality of features space. At last, we applied Gentle-Adaboost classifier to grade different level gliomas. The experiment results show that the prominent performance of our approach on our image database. As an increase in image database size, we consider that our radiomic prediction model will be as an auxiliary tool which can provide malignant and benign information of glioma to improve the accuracy efficiently.