An improved whale optimization algorithm-based radial neural network for multi-grade brain tumor classification

Nowadays, brain tumor has become deadly disease. Therefore, early diagnosis can prevent the severity of the next stage. The existing methods are inefficient in predicting the optimal cluster center (CC) in radial basis neural network (RBNN), which leads to inaccuracy. To overcome this issue, we proposed an improved whale optimization algorithm (IWOA) of RBNN to maximize the convergence speed and accuracy. Initially, the MRI input images are fed to preprocessing steps. Then, the image segmentation is carried out by fuzzy-c means (FCM) clustering for identifying the tumor region. These tumor and non-tumor images undergo a feature extraction process utilizing principle component analysis (PCA), mean, entropy, and wavelet transform. The obtained feature vector is fed to the RBNN layer. It needs an optimal CC, which can be considered by using the newly proposed IWOA. RBNN classifies the abnormality of the brain into brain tumor, inflammatory disease, stroke, and degenerative. Its performance is tested on three datasets. Based on the report of the evaluation, the proposed FCM-IWOA-RBNN gives high accuracy.