An automated and risk free WHO grading of glioma from MRI images using CNN

Glioma is among aggressive and common brain tumors, with a low survival rate, in its highest grade. Invasive methods, i.e., biopsy and spinal tap are clinically used to determine the grades of glioma. Depending upon the findings of these methods, treatment is planned to improve the life expectancy of the controls. Magnetic resonance imaging (MRI), the most widely used medical imaging modality to diagnose a brain tumor, is producing a huge volume of MRI data. A reliable, automatic, and noninvasive method of glioma grading are always required as an alternative to these invasive methods. In this research, a model has been proposed using Convolutional Neural Networks to classify low and high-grade glioma. A locally organized dataset, developed in the Department of Radiology (Diagnostics), Bahawal Victoria Hospital, Bahawalpur, Pakistan has been used for research and experiments. Additionally, results have also been validated on a publicly available benchmarked dataset, i.e., BraTS-2017. The proposed method demonstrated significant achievement in terms of classification rates, i.e., the accuracy of 98.93% (for low-grade glioma) and 98.12% (for high-grade glioma). Experimental results proved that the proposed model is accurate (98.52%) and is efficient in glioma grade identification.

Automated summary

In this research, a model using Convolutional Neural Networks for glioma grading is proposed. The model is validated using a locally organized dataset and a publicly available benchmarked dataset. Experimental results demonstrate high accuracy and efficiency of the proposed model in glioma grade identification.