Classification of Brain Tumor using Manta Ray Foraging Optimization-based DeepCNN Classifier

Abstract

The abnormal cell growth in the brain is the brain tumor. Early diagnosis of brain tumor helps in preventing cancer from advancing to the next level. The major concern in the brain tumor diagnosis is accuracy. In this research, a brain tumor classification method is developed using the Manta Ray Foraging Optimization-based Deep Convolutional Neural Network (MROA-based DeepCNN) algorithm. The preprocessing of Magnetic Resonance (MR) images is done with the Gaussian filter for removing the artifacts in the MR image. The cellular automata and rough set theory are used for segmenting the cancerous region from the preprocessed image. The CNN feature extracted from the segmented output is given as the input to the DeepCNN classifier. The brain tumor is classified into benign, core, edema, and malignant tumor using the DeepCNN classifier, which is trained using the MRFO algorithm. The developed MROA-based DeepCNN method is evaluated using metrics like sensitivity, accuracy, and specificity. While comparing with the existing brain tumor classification methods, the developed MROA-based DeepCNN method obtained a maximum accuracy of 0.9899, maximum sensitivity of 0.8316, maximum specificity of 0.9899.

References