2D MRI registration using glowworm swarm optimization with partial opposition-based learning for brain tumor progression

Magnetic resonance imaging (MRI) registration is important in detection, diagnosis, treatment planning, determining radiographic progression, functional studies, computer-guided surgeries, and computer-guided therapies. The registration process is the way to solve the correspondence problem between features on MRI scans acquired at different time-points to study the changes while analyzing the brain tumor progression. Registration method generally requires a search strategy (optimizer) to search the transformation parameters of the registration to optimize some similarity metric between images. Metaheuristic algorithms are becoming more popular recently for image registration. In this paper, at the outset, a metaheuristic algorithm, namely glowworm swarm optimization (GSO), is improved by incorporating partial opposition-based learning (POBL) strategy. The improved GSO is applied to register the pre- and post-treatment MR images for brain tumor progression. A comparative study has been made with basic GSO, GSO with generalized opposition-based learning (GOBL-GSO), and existing particle swarm optimizer (PSO)-based registration method. The experimental results demonstrate that the proposed method has an extremely higher statistical significance in performance than others in brain MRI registration.