冠脉CT血管成像(CCTA)能够在高风险人群中有效排除冠心病，但是容易受心脏节律性搏动影响产生运动伪影，使图像模糊，影响诊断。Wasserstein生成式对抗网络(WGAN)是一种新的深度学习模型，其中的生成网络根据训练数据生成图像，判别网络通过判断生成图像与参考图像特征是否相符，重复迭代提高图像清晰度，可用于从模糊图像生成清晰图像。前期研究发现深度学习可识别冠脉钙化斑块伪影，回溯原斑块。本课题拟基于WGAN从模糊CCTA图像生成无伪影的清晰图像。纳入1,200例受检者的CCTA图像，其中80%为训练集，20%为测试集。训练并优化WGAN网络结构和参数，建立从能去除图像运动伪影的深度学习模型。最后前瞻性纳入100例有CCTA和冠脉造影的病例验证此科学方法产生图像对冠脉狭窄的诊断准确性。本研究将从理论上阐明WGAN对CCTA去伪影的科学价值，促进现有CT对冠心病诊断的临床应用。

Coronary CT angiography (CCTA) effectively rules out coronary artery disease in high-risk population. But CCTA is unavoidably influenced by motion artefact because of periodical cardiac beat. Blurred CCTA images eventually decreases diagnostic accuracy. Wasserstein generative adversarial network (WGAN) is a state-of-the-art deep learning model. Generative network produces images from the training data. Discriminative network judges the agreement between generated and reference images relying on internal image features, thus repeatedly and iteratively improves image resolution and clarity. WGAN is possible to generate clear image from blurred image. The previous study showed that deep learning recognized features of motion artefact of coronary calcified plaque, thus was possible to recall the original plaque. In this study, we hypothesize that clear CCTA image without motion artefact could be generated from original blurred image by optimized WGAN with technical and image features of CCTA. We plan to retrospectively collect 1,200 cases of CCTA examinee, including 80% in the training dataset and 20% in the test dataset. Subsequently, we intend to establish a WGAN model and adjust its architecture and parameters for CCTA from blurred image to clear image. Finally, 100 patients who have CCTA and invasive coronary angiography will be prospectively included to validate the diagnostic accuracy on coronary stenosis using images generated by this scientific method. This study will theoretically clarify the scientific value of WGAN on CCTA motion artefact removal, and contribute to the application of current CT equipment on diagnosis of coronary artery disease.