Fractional probabilistic fuzzy clustering and optimization based brain tumor segmentation and classification

Brain tumor detection concentrates on the automatic segmentation and classification of the brain tumors using the MRI brain images. The proposed work includes pre-processing, segmentation, feature extraction and classification. In the pre-processing step the RGB image is subjected to hue colour transformation and then thresholding and Morphological operations are performed. The automatic segmentation is initiated using Fractional probabilistic Fuzzy C Means (Fr-pFCM) by combining the fractional theory and probabilistic Fuzzy Clustering (probabilistic FCM) to enable the highly accurate segments. The features are extracted from the segments for which the descriptors, Empirical Mode Decomposition (EMD), Local Directional Pattern (LDP), wavelet transform, and theoretic information measures are employed. The final classification is done using the Whale-Cat Swarm Optimization based Deep Belief Network (WCSO-DBN). The experimentation using the images from the BRATS database outperformed the existing classification methods with higher accuracy, sensitivity, and specificity of 0.923, 0.95, and 0.96, respectively.