The appearance of generative adversarial networks (GAN) provides a new approach and framework for computer vision. Compared with traditional machine learning algorithms, GAN works via adversarial training concept and is more powerful in both feature learning and representation. GAN also exhibits some problems, such as non-convergence, model collapse, and uncontrollability due to high degree of freedom. How to improve the theory of GAN and apply it to computer-vision-related tasks have now attracted much research efforts. In this paper, recently proposed GAN models and their applications in computer vision are systematically reviewed. In particular, we firstly survey the history and development of generative algorithms, the mechanism of GAN, its fundamental network structures, and theoretical analysis of the original GAN. Classical GAN algorithms are then compared comprehensively in terms of the mechanism, visual results of generated samples, and Frechet Inception Distance. These networks are further evaluated from network construction, performance, and applicability aspects by extensive experiments conducted over public datasets. After that, several typical applications of GAN in computer vision, including high-quality samples generation, style transfer, and image translation, are examined. Finally, some existing problems of GAN are summarized and discussed and potential future research topics are forecasted. INDEX TERMS Deep learning, generative adversarial networks (GAN), computer vision (CV), image generation, style transfer, image inpainting.

No abstract

Background: This study was designed to explore the progression patterns of IDHwildtype glioblastoma (GBM) at first recurrence after chemoradiotherapy. Methods: Records from 247 patients who underwent progression after diagnosis of IDHwildtype GBM was retrospectively reviewed. Progression patterns were classified as either local, distant, subependymal or leptomeningeal dissemination based on the preoperative and serial postoperative radiographic images. The clinical and molecular characteristics of different progression patterns were analyzed. Results: A total of 186 (75.3%) patients had local progression, 15 (6.1%) patients had distant progression, 33 (13.3%) patients had subependymal dissemination, and 13 (5.3%) patients had leptomeningeal dissemination. The most favorable survival occurred in patients with local progression, while no significant difference of survival was found among patients with distant progression, subependymal or leptomeningeal dissemination who were thereby reclassified into non-local group. Multivariable analysis showed that chemotherapy was a protective factor for non-local progression, while gender of male, subventricular zone (SVZ) involvement and O 6-methylguanine-DNAmethyltransferase (MGMT) promoter methylation were confirmed as risk factors for non-local progression (P < 0.05). Based on the factors screened by multivariable analysis, a nomogram was constructed which conferred high accuracy in predicting non-local progression. Patients in non-local group could be divided into long-and shortterm survivors who differed in the rates of SVZ involvement, MGMT promoter methylation and reirradiation (P < 0.05), and a nomogram integrating these factors showed high accuracy in predicting long-term survivors. Conclusion: Patients harboring different progression patterns conferred distinct clinical and molecular characteristics. Our nomograms could provide theoretical references for physicians to make more personalized and precise treatment decisions.