{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/how-can-we-make-gan-perform-better-in-single","title":"How Can We Make GAN Perform Better in Single Medical Image Super-Resolution? A Lesion Focused Multi-Scale Approach","arxiv_id":"1901.03419","date":"2019-01-10","proceeding":null,"authors":["Jin Zhu","Guang Yang","Pietro Lio"],"abstract":"Single image super-resolution (SISR) is of great importance as a low-level\ncomputer vision task. The fast development of Generative Adversarial Network\n(GAN) based deep learning architectures realises an efficient and effective\nSISR to boost the spatial resolution of natural images captured by digital\ncameras. However, the SISR for medical images is still a very challenging\nproblem. This is due to (1) compared to natural images, in general, medical\nimages have lower signal to noise ratios, (2) GAN based models pre-trained on\nnatural images may synthesise unrealistic patterns in medical images which\ncould affect the clinical interpretation and diagnosis, and (3) the vanilla GAN\narchitecture may suffer from unstable training and collapse mode that can also\naffect the SISR results. In this paper, we propose a novel lesion focused SR\n(LFSR) method, which incorporates GAN to achieve perceptually realistic SISR\nresults for brain tumour MRI images. More importantly, we test and make\ncomparison using recently developed GAN variations, e.g., Wasserstein GAN\n(WGAN) and WGAN with Gradient Penalty (WGAN-GP), and propose a novel\nmulti-scale GAN (MS-GAN), to achieve a more stabilised and efficient training\nand improved perceptual quality of the super-resolved results. Based on both\nquantitative evaluations and our designed mean opinion score, the proposed LFSR\ncoupled with MS-GAN has performed better in terms of both perceptual quality\nand efficiency.","url_abs":"http://arxiv.org/abs/1901.03419v1","url_pdf":"http://arxiv.org/pdf/1901.03419v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"how-can-we-make-gan-perform-better-in-single","repo_url":"https://github.com/GinZhu/MIASSR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"how-can-we-make-gan-perform-better-in-single","repo_url":"https://github.com/GinZhu/RDST","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":null,"task_name":"Generative Adversarial Network"},{"task_slug":"image-super-resolution","task_name":"Image Super-Resolution"},{"task_slug":"super-resolution","task_name":"Super-Resolution"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"wgan","method_name":"WGAN"},{"method_slug":"wgan-gp","method_name":"WGAN GP"},{"method_slug":"wgan-gp-loss","method_name":"WGAN-GP Loss"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}