Wu Xiuxiu, Xiao Shan, Zhang Yu
Annu Int Conf IEEE Eng Med Biol Soc. 2014;2014:2444-7. doi: 10.1109/EMBC.2014.6944116.
Lung 4D-CT plays an important role in lung cancer radiotherapy for tumor localization and treatment planning. In lung 4D-CT data, the resolution in the slice direction is often much lower than the in-plane resolution. For multi-plane display, isotropic resolution is necessary, but the commonly used interpolation operation will blur the images. In this paper, we present a registration based method for super resolution enhancement of the 4D-CT multi-plane images. Our working premise is that the low-resolution images of different phases at the corresponding position can be regarded as input "frames" to reconstruct high resolution images. First, we employ the Demons registration algorithm to estimate the motion field between different "frames". Then, the projections onto convex sets (POCS) approach is employed to reconstruction high-resolution lung images. We show that our method can get clearer lung images and enhance image structure, compared with the cubic spline interpolation and back projection method.
肺部4D-CT在肺癌放射治疗的肿瘤定位和治疗计划中起着重要作用。在肺部4D-CT数据中，切片方向的分辨率通常远低于平面内分辨率。对于多平面显示，各向同性分辨率是必要的，但常用的插值操作会使图像模糊。在本文中，我们提出了一种基于配准的方法，用于增强4D-CT多平面图像的超分辨率。我们的工作前提是，将相应位置不同相位的低分辨率图像视为输入“帧”来重建高分辨率图像。首先，我们采用Demons配准算法来估计不同“帧”之间的运动场。然后，采用凸集投影（POCS）方法来重建高分辨率肺部图像。我们表明，与三次样条插值和反投影方法相比，我们的方法可以获得更清晰的肺部图像并增强图像结构。
