Fractional Aquila spider monkey optimization based deep learning network for classification of brain tumor

The tumor in the brain is a serious disease that causes death in humans. Various imaging modalities are utilized for identifying tumors, but the huge data produced by magnetic resonance imaging (MRI) can affect the manual classification of brain tumors. Thus, an automatic technique to categorize tumors is highly desirable. A novel optimization-driven model is devised to classify brain tumors. The hybrid segmentation approach is employed by fusing the U-Net and channel-wise feature pyramid network for medicine (CFPNet-M) model with Tanimoto similarity. The training of the segmentation network is performed with Aquila spider monkey optimization (ASMO), which is the integration of the spider monkey optimization (SMO) and Aquila optimizer (AO). The hybrid segmentation model precisely segments and categorizes the benign and malignant tumor samples. The proposed segmentation model is adapted to obtain the MRI segment. Moreover, an attempt is made to train the SqueezeNet model for classifying the tumors into four grades. The weights of SqueezeNet are optimally tuned using devised technique, namely fractional Aquila spider monkey optimization (FASMO), which is obtained by integrating SMO, AO, and fractional calculus (FC). The proposed FASMO-based SqueezeNet offered enhanced efficiency with 92.2% testing accuracy, 94.3% sensitivity, 90.8% specificity and 0.089 prediction error.