Brain Tumor Detection Based on Hybrid Deep Neural Network in MRI by Wind Driven Water Wave Optimization

Abstract

Image processing techniques are helpful in several applications mainly in the medical field for the examination of computerized disease. The medical imaging method named Magnetic Resonance Imaging (MRI) is broadly used to record the information about brain tumor for the future clinical study and research. Automatic evaluation approaches are required for the assessment of brain MRI due to its various modality and complexity. This paper presents a novel approach to support the examination of brain MRI. Also, it extracts the edema sector and tumor core from the image using Wind Driven Water Wave Optimization (WDWWO), entropy value and contour based segmentation. Deep Neural Network based Elephant Herding Optimization (DNN-EHO) is proposed for the classification. The proposed work is implemented in MATLAB 2018 platform and the severity of tumor is identified. Further, the WDWWO approach can be validated on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2012 dataset challenge on BRATS (Brain Tumor Segmentation) and attained better outcomes in terms of accuracy, recall, specificity, precision, and F-measure. The testing process for this proposed method is carried out with the T1C, Flair, and T2 modalities. Further, with the BRATS 2012 challenge dataset, WDWWO assisted tumor detection framework is validated and attained better values for F-measure, precision, specificity, sensitivity, and accuracy.