Stroke is a leading cause of death and disability worldwide, with ischemic stroke most common. Brain CT images of patients who have suffered ischemic stroke show obvious features. Hence, brain CT images are often used to determine whether a patient has suffered an ischemic stroke. In recent years, there has been rapid development of software, hardware, and artificial intelligence (AI) applications. To speed up the diagnostic process, medical images can be labeled using AI technology. As medical images are difficult to acquire, we applied Generative Adversarial Network (GAN) deep learning algorithm. Through the network of generators and discriminators in the GAN model, there is continuous learning with improved parameters from a small amount of data. With this model, the best parameters are determined to generate labeled brain CT images that closely match those manually labeled by physicians. The method proposed in this paper can be used to precisely label the regions of ischemic stroke. In addition, we provide a platform for doctors to label brain CT images. We hope that the proposed method can effectively assist physicians in studying brain CT images, thereby shortening the diagnostic time, accelerating the treatment process, and providing an understanding of a patient's condition in the shortest amount of time.