Optimization empowered hierarchical residual VGGNet19 network for multi-class brain tumour classification

Brain tumour is a fatal disease and its diagnosis is a difficult procedure for radiologists due to the heterogeneous behaviour of tumour cells. Diagnosing brain tumour using MRI at early stage is essential by attaining higher accuracy with minimum error function. Based on this insight, the goal of the proposed research is to create a deep learning architecture that will aid in the automatic detection of brain tumour utilising two-dimensional MRI slices. The proposedSquare Array Filtering(SAF) approach is applied to the acquired input image in order to remove such noisy contents. To obtain a noiseless image, an array in square grid format is generated to update the missing pixel values of an image using SAF. This research employs the Kernel K-means clustering ((KC)-C-2) model for segmenting the injured region, with the weighted mean enhancement methodology efficiently removing noise from the image and the (KC)-C-2 method isolating the infected areas. The proposed Optimization empowered Hierarchical Residual VGGNet19 (HR-VGGNet19) model is designed to explore the discriminative information with the help of convolution layer employed in it. The proposed model combines both the low level and high-level features with the help of HR-VGGNet19 network and obtains the output class. The performance metric shows the proposed methodology outperforms better than the state-of-art methods. The proposed methodology is implemented in the working platform of MATLAB in terms of evaluation metrics like accuracy, sensitivity, specificity, PPV, NPV, FPR, FNR and FDR.