Evaluating the Effectiveness of Training a Generative Adversarial Network Through Transfer Learning for the Purpose of Completing CT Images