Optimizing Convolutional Neural Network by Hybridized Elephant Herding Optimization Algorithm for Magnetic Resonance Image Classification of Glioma Brain Tumor Grade

Gliomas belong to the group of the most frequent types of brain tumors. For this specific type of brain tumors, in its beginning stages, it is extremely complex to get the exact diagnosis. Even with the works from the most experienced doctors, it will not be possible without magnetic resonance imaging, which aids to make the diagnosis of brain tumors. In order to create classification of the images, to where the class of glioma belongs to, for achieving superior performance, convolutional neural networks can be used. For achieving high-level accuracy on the image classification, the convolutional network hyperparameters’ calibrations must reach a very accurate response of high accuracy results and this task proves to take up a lot of computational time and energy. Proceeding with the proposed solution, in this scientific research paper a metaheuristic method has been proposed to automatically search and target the near-optimal values of convolutional neural network hyperparameters based on hybridized version of elephant herding optimization swarm intelligence metaheuristics. The hybridized elephant herding optimization has been incorporated for convolutional neural network hyperparameters’ tuning to develop a system for automatic and instantaneous image classification of glioma brain tumors grades from the magnetic resonance imaging. Comparative analysis was performed with other methods tested on the same problem instance an results proved superiority of approach proposed in this paper.