Generative Adversarial Networks for Medical Image Super-resolution

Super-resolution (SR) techniques are very useful in enhancing low resolution images. This becomes even more effective when the clinicians and radiologists need to detect tiny bone fractures in some low-resolution medical images such as X-Rays. In this short paper, the application of new deep learning single-image SR techniques to medical bone X-Ray images have been investigated. The quality of the results, when applied to plain hand X-Rays, are assessed based on peak signal-to-noise ratio and mean opinion score (MOS) and the superiority of generative adversarial networks (GANs), particularly in terms of MOS, has been verified for such applications.