Tumor bagging: a novel framework for brain tumor segmentation using metaheuristic optimization algorithms

Brain tumor segmentation is a challenging research problem and several methods in literature have been suggested for addressing the same. In this paper, we propose a novel framework called Tumor Bagging whose objective is to enhance the performance of brain tumor segmentation by combining more than one segmentation methods based on Multilayer Perceptron (MLP). For this purpose, three metaheuristic optimization algorithms viz. Gray Wolf Optimizer, Artificial Electric Field Optimization Algorithm and Spider Monkey Optimization have been exploited for learning the network parameters of MLP. The results from these three models are further combined based on majority voting method. We have exploited three different magnetic resonance modalities i.e. Fluid-Attenuated Inversion Recovery (FLAIR), contrast-enhanced T1, and T2 for experiments. Three brain tumor regions i.e. complete tumor, enhancing tumor, and tumor core are segmented. The advantage of the proposed method is its simplicity as well as it gives significant and improved performance using the bagging approach on the publicly available and benchmark BRATS dataset. Dice Similarity Coefficient (DSC) is a performance measure which combines positive predictive value and sensitivity. We have achieved a DSC score of more than 92% for detection of complete tumor region in high-grade as well as low-grade glioma subjects that is better than several state-of-the-art methods.

Automated summary

The paper introduces a novel framework called Tumor Bagging for enhancing brain tumor segmentation by combining multiple segmentation methods and utilizing different optimization algorithms. Experimental results demonstrate significant and improved performance on the BRATS dataset, achieving high accuracy in segmenting different brain tumor regions.