Efficient Segmentation of Tumor and Edema MR Images Using Optimized FFNN Algorithm

Abstract

AbstractBrain tumor categorization and segmentation play a crucial in the existing healthcare domain. In this research study, an effective technique for brain magnetic resonance imaging (MRI) image segmentation and classification is demonstrated in order to improve accurate brain tumor segmentation. The anticipated brain tumor segmentation technique includes five phases, namely preprocessing, filtering, feature extraction, and classification along with segmentation. In the preprocessing stage, from the MRI database, the input MRI image is primarily fetched, and additionally, it is exposed to the skull stripping in order to remove the unwanted zone from the image. Skull stripped image is refined by utilizing a filter recognized as the adaptive median filter. However, the constituents of discrete wavelet transform (DWT), probability, skewness, and kurtosis are detached from the filtered images. Cuckoo search optimized feedforward neural network (FFNN) classifier categorizes the brain images like normal and also abnormal concerning the extracted features. Finally, the gray matter (GM), white matter (WM) along with cerebrospinal fluid (CSF) are apportioned as of the normal images through k-means clustering, and later, the cancer is segmented by utilizing EM algorithm, and the edema region is apportioned by utilizing watershed algorithm from the anomalous images, respectively. Henceforth, the outcomes will be investigated for exhibiting the accomplishment of the recommended classification and also segmentation techniques.KeywordsSegmentationFeature extractionOptimizationFilteringSkull strippingClassification