An integrated optimized hybrid intensity modeled brain tumor image segmentation using artificial bee colony algorithm

Brain magnetic resonance (MR) images are very significant in early detection and diagnosis of the brain tumor treatment. In order to improve the efficiency and accuracy of the brain tumor segmentation in MR images, an integrated hybrid segmentation approach is proposed. This approach is developed by integrating the features of local independent projection based classification (LIPC), partitioning around the medoids (PAM) and enhanced fuzzy c- means (EFCM) approach that is employed for the determination of the cluster centers that classify MR images accurately based on independent projections of the tumor cells. This segmentation approach is considered as classification problem. The classification accuracy is improved by incorporating additional contextual features for the tumor cells. The segmented image from the hybrid method may have some healthy cells included in the tumor regions due to tumor mass effect. Thus an improved intensity modeling is proposed to eliminate the tumor mass effect and also to improve the accuracy in classification. The above work utilizes the artificial bee colony (ABC) algorithm for the improvement in the segmentation of tumor cells below the surface area. Here, the tumor regions are clustered into colonies and each colony is enhanced to a solution. The colonies are combined with the synthetic data of the images. The ABC technique enables the segmentation even in deeper layers by partitioning the tumor images around the lesions that makes the proposed system to better result in efficient detection and segmentation of brain tumor from the MR images.