Minimally parametrized segmentation framework with dual metaheuristic optimisation algorithms and FCM for detection of anomalies in MR brain images

Background: Early prognosis of a brain tumour may offer better life expectancy. Magnetic Resonance Imaging (MRI) coupled with an efficient machine learning segmentation technique has proven to be a reliable way of assessing tumours. In addition to the segmentation, the image is needed to optimise to achieve the desired results. In many cases, single-stage optimisation could not complete the search target owing to algorithm-specific limitations. To overcome this hindrance, the dual metaheuristic optimisation technique is widely used to detect tumour affected tissues. Aim: This research emphasises brain tumour region detection using Fuzzy C-Means (FCM) clustering techniques and the segmented output enhancement using two different optimisation techniques, namely, Artificial Bee Colony (ABC) and the JAYA algorithm. Methods: This methodology first deploys the FCM clustering technique to segment the tumour region in the MRI. Then, the initial stage of optimisation is done using the ABC algorithm with the help of texture features extracted from the segmented image through the Gray Level Co-occurrence Matrix (GLCM) technique. Lastly, a novel JAYA algorithm is deployed for the second stage of optimisation to provide precise segmentation with the support of global and local best solutions. Result and Conclusion: The proposed framework delivers high accuracy in tumour detection. Besides, it has been proven by renowned evaluation metrics, such as Tanimoto Coefficient Index, and Dice Coefficient Index, which are up to 70.12% and 82.56%, respectively, competing with the contemporary methods used for the evaluations of MR brain images.