Self-guided Genetic Algorithm for Limited View Tomography

Computed tomography (CT) is a noninvasive method to know the objects’ internal structure. Due to its noninvasive nature, CT methods are widely used in medical diagnosis. However, CT methods produce high radiation doses. Therefore, limited data reconstruction methods require for reducing radiation dose and cost per scan. The proposed manuscript proposes a self-guided genetic algorithm for limited view tomography. The algorithm uses two methods, self-guided crossover and self-guided mutation operation, for finding the optimal solution. The proposed algorithm also uses adaptive crossover and mutation rate, improving convergence rate and reduce the chance of local stagnation. Experimental results reveal that the proposed algorithm produces satisfactory results with limited-angle projection data. The algorithm also outperforms the other state-of-the-art reconstruction algorithm for limited view data or sparse data.