The invention discloses a U-net based generation
adversarial network based digital substration
angiography DSA imaging method and device. The method comprises the following steps: firstly, obtaining aplurality of groups of
radiography frames and
original data corresponding to the subtraction frames; secondly, establishing a U-shaped structure convolutional network, coding and extracting featuresof different scales, and decoding and recovering corresponding features by using a skill
connected object, wherein the network inputs a
radiography frame and outputs a corresponding subtraction frameto reduce the dependence of DSA generation on a background frame so as to remove
motion artifacts generated by motion of a patient; then, by means of generative adversarial training, alternately training the generator and the
discriminator, and enhancing the quality of generated digital
blood vessel subtraction. According to the method, the
motion artifacts in the digital substration
angiography (DSA) can be effectively removed, the
data quality can meet the requirements of clinical analysis, diagnosis and the like, the DSA
imaging quality is improved, and the influence caused by motion of a patient is reduced.