A Multimodal Biomedical Image Registration Method Based on an Improved Genetic Algorithm Inspired by Hybrid Breeding.

Image Registration(IR) has been widely applied in biomedical image processing. It is the process of finding an optimal geometric transformation to align two images, which could be defined as a parameter optimization issue. Genetic Algorithm(GA) is one of the most efficient methods for solving complex optimization problems and it has been applied to the real-coding IR problem. However, the classical GA suffers from premature convergence and is easy to fall into local optimum. Inspired by heterosis, which is a common phenomenon in biology, this study proposes an improved GA. By artificially simulating the breeding process of Chinese three-line hybrid rice, known as a successful application of heterosis, the original crossover and mutation mechanisms of GA are improved, and a dynamic diversity controller is designed. This study conducts several multimodal biomedical IR experiments to compare the improved GA with state-of-the-art IR methods, the results show that the proposed method outperforms the others in most scenarios with faster convergence speed and greater robustness.