论文部分内容阅读
多层螺旋CT在临床影像诊断中被广泛应用,由于X线是一种电离辐射,对人体有一定伤害,高剂量的X线辐射会增加人体罹患癌症等疾病的风险,因此在保证诊断效果的前提下尽可能地降低X射线剂量是当前研究的热点问题。考虑到螺旋扫描方式在纵向采样具有不一致的特性,本文提出了一种基于三维字典学习的统计迭代重建算法用于低剂量螺旋CT重建,同时在迭代框架中引入纵向TV约束以改变螺旋重建图像的噪声分布特性。为了保证重建结果的精确度,经典的基于距离驱动的前向投影和反投影被用于迭代重建框架。同时为了提升计算速度。本文核心算法采用GPU实现,有序子集和Nesterov方法也用于加速迭代收敛过程。本文提出的方法能够有效的提高图像质量,对比度噪声比指标和医学从业人员的主观评价指标验证了本文算法的优越性。
Multi-slice spiral CT is widely used in clinical imaging diagnosis. Because X-ray is an ionizing radiation, it will cause some harm to the human body. High dose of X-ray will increase the risk of cancer and other diseases. Therefore, As far as possible under the premise of reducing X-ray dose is currently a hot issue. Considering the inconsistent characteristics of helical scanning in longitudinal sampling, a statistical iterative reconstruction algorithm based on three-dimensional dictionary learning is proposed for low-dose spiral CT reconstruction. Meanwhile, a longitudinal TV constraint is introduced in the iterative framework to change the helical reconstructed image Noise distribution characteristics. To ensure the accuracy of reconstruction results, classical distance-driven forward projection and back-projection are used to iteratively reconstruct the frame. At the same time in order to improve the calculation speed. The core algorithm in this paper uses GPU, the ordered subset and Nesterov method are also used to accelerate the iterative convergence process. The proposed method can effectively improve the image quality. The contrast-to-noise ratio and the subjective evaluation index of medical practitioners validate the superiority of the proposed algorithm.