Integrated Hybrid Segmentation to Overcome Problems in Brain Tumor Segmentation

Background/Objective: To improve the performance of the segmentation of Brain tumor images by introducing integrated hybrid segmentation approach. Methods/Statistical Analysis: An innovative integrated approach combining a set of features of some efficient algorithms is proposed in this paper. The proposed algorithm known as integrated hybrid segmentation approach uses the features that classify the MRI images based on the local independent projections. The hybrid PAM and enhanced possibilistic fuzzy C-means approach is used in the partition and determination of the cluster centers. Findings: The problem of false segmentation and segmentation with low accuracy that prevails in the brain image segmentation can be overcome by using this approach. The classified images are used as training data using which different dictionaries are constructed with all classes. The testing samples are projected to the dictionaries and reconstructed using local anchor embedding approach. Thus the proposed integrated algorithm improves the segmentation process of brain tumor images considerably. Applications/Improvements: Improved segmentation performance is required for accurate detection of brain tumor. The integrated hybrid segmentation approach has better segmentation accuracy than other segmentation approaches with 6% improvement which is very significant.