M3BTCNet: multi model brain tumor classification using metaheuristic deep neural network features optimization

Brain tumor is an active research topic in the area of medical imaging due to only 36% survival rate. The malignant tumor is a more dangerous type of tumor. The recent facts and figures show that 700,000 American are living in brain tumor and from them, 30.25% tumors are malignant. The diagnosis at the initial stage can help to minimize the human mortality rate due to brain tumor. Several researchers of computer vision proposed computer-aided diagnosis systems for this purpose but they faced several issues such as low accuracy and higher computational time. In this work, an end-to-end optimized deep learning system for multimodal brain tumor classification is proposed. The BRATS datasets are utilized for evaluation. The contrast is enhanced at the very first step by using hybrid division histogram equalization along ant colony optimization approach and train a newly design nine-layered CNN model. The features are extracted from second fully connected layer and optimized through differential evolution and mouth flame optimization. The output of both methods is fused using matrix length approach and passed in multi-class support vector machine (MC-SVM). In the experimental process, suggested method achieved an accuracy of 99.06, 98.76, 98.18 and 94.6%, on BRATS 2013, BRATS 2015, BRATS 2017 and BRATS 2018, respectively. Comparing results with existing techniques shows the superiority of proposed technique.