Absolute Structure Threshold Segmentation Technique Based Brain Tumor Detection Using Deep Belief Convolution Neural Classifier

Brain tumors are caused by abnormal cells developing in the human brain. The incidence of malignant brain tumors is relatively high and significantly influences humans and society. Magnetic Resonance Imaging (MRI) is an excellent non-invasive technique that produces high-quality brain images without damage. And it makes an adequate diagnosis and is considered the primary technical treatment. This type of method of tumor identification has some problems, there are less efficient for complex tumor stages and increases computation time, and segmentation is an inaccurate and unreliable result. To tackle this problem, this paper proposes Absolute Structure Threshold Segmentation Technique (ASTST) based on Deep Belief Convolution Neural Classifier (DBCNC) using Softmax activation function for brain tumor classification. The proposed method initially starts with the preprocessing step supported by completing the Gaussian and Bilateral Filter(GBF) using the brain images to remove the Noise, enhance the image size and color contrast level and enhance the frequency of the images to find the tumor-affected area. After preprocessing image is trained into Absolute Gabor with Canny Edge Selection (AGCES) technique to identify the edges without affecting the image quality. Then the Similarity Scaling Shapes Feature Selection (S3FS) method is used to analyze the most delicate features of brain tumors relatively to find the dimension to improve the accuracy. Based on the feature selection, the proposed DBCNC algorithm classifies the brain tumor as malignant or Normal. The proposed method improves prediction accuracy, sensitivity, specificity, and f-measure and minimizes time complexity and false rate. &nbsp