一种基于深度学习的心脏DSA影像增强技术的建立

Establishment of a deep learning based cardiac DSA image enhancement technology

目的探讨一种新型的心脏DSA图像血管增强技术。方法首先收集了1089幅心脏DSA原始影像,其中部分影像存在一些问题,例如噪声、颜色偏暗等;我们采用卷积神经网络结合传统的医学图像处理技术进行心脏DSA图像增强。结果经过卷积神经网络处理后,心脏DSA图像变得更清晰,血管影像的亮度及对比度都得到明显增强,对比度增强2.4倍。结论采用深度学习技术进行心脏DSA图像的血管增强,具有效果好、速度快、稳定性高、且适用范围大的特点,更适合心脏DSA设备采用。

Objective To develop a new methodology for Cardiac DSA image enhancement.Methods 1089 original Cardiac DSA images were collected,part of which has quality problems,such as noise and uneven brightness.A convolutional neural network combined with traditional medical image processing technology was implemented to enhance the Car-diac DSA images for vessel illustration.Results After being processed by the deep learning network,the Cardiac DSA images become clearer,brighter,and the contrast of blood vessels were significantly enhanced,Contrast enhancement was 2.4 times.Conclusion Deep learning has a great advantage for medical image processing.Our methodology for the enhancement of Cardiac DSA images has many advantages,including capability,effectiveness,robustness,and universality,which make it adaptable for Cardiac DSA equipment.