Ensemble based clustering for brain tumor segmentation and super learner based classification

Abstract

In the field of medical image analysis, Computer-Aided Diagnosis newline(CAD) attracts more attention in recent years. The automatic segmentation of newlinetumor from Brain MRI and classification is one of the major research problems newlinein the medical field. The brain is composed of gray and white matter where as newlinethe brain tumor is the unusual growth of cells in a particular region. The brain newlinetumor is one of the major deaths defying diseases. The detection of tumors in the newlineearly stages will increase the patientand#8223;s survival rate. The manual segmentation of newlinetumor from brain MRI is time-consuming process and may be prone to error. So newlinethe automatic segmentation and classification of brain tumor plays a key role in newlinemedical applications. Machine learning, deep learning based algorithms make newlineaccurate predictions from input data without human intervention. So machine, newlinedeep learning based algorithms act as a base for CAD. newlineSegmentation is used to predict the Region of Interest (ROI) from the newlinebrain image and the classification technique is used to classify the severity of the newlinedisease. In the segmentation process, the tissue regions of the brain image are newlinedivided into clusters based on the relevancy between the intensity of pixels. newlineThen the class of tumor based on severity is achieved by classification newlinealgorithms. There are several clustering and classification machine learning newlinealgorithms available to automate the process of segmentation and classification. newlineBut each algorithm has its own advantage and disadvantage. Even though deep newlinelearning algorithms provide accurate results, they need more data samples. newlineAnother disadvantage of deep learning is opaqueness in results. The main newlineobjective of the proposed work is to create an ensemble model for brain tumor newlinesegmentation and classification. The ensemble is the process of combining the newlinepredictions from different algorithms by this we will get improved results newlinecompared to a single algorithm. In the proposed work three different types of newlineMRI such as T1-weighted MRI, T2-Weighted MRI and FLAIR MRI are