Super-Resolution Reconstruction of CT Images Based on Generative Adversarial Networks

As Computerized Tomography (CT) images are widely used in medical diagnosis, obtaining high-resolution images is crucial for improving diagnostic accuracy. Due to current limitations in equipment and technology, and considering the radiation damage to the human body caused by obtaining high-resolution images, the CT images currently generated have relatively low resolution. This paper introduces the use of a Generative Adversarial Network (GAN) for super-resolution reconstruction of CT images. At the algorithm level, the acquired low-resolution images are transformed into high-resolution images, thereby improving the visual quality of CT scans while maintaining a low radiation dose. Compared with traditional bicubic interpolation, Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) are improved by 22.2% and 19.0%, respectively. Considering the low contrast and limited features in CT images, the introduction of dense residual blocks enhances the existing algorithm, resulting in a 12.3% improvement of PSNR and 1.3% improvement of SSIM respectively, which indicating that the improved SRGAN algorithm is more similar to the original high-resolution image and has less distortion, which further proves the superiority of the algorithm.