Motion correction for MR cystography by an image processing approach.
Motion correction for MR cystography by an image processing approach.
复制标题
通过图像处理方法进行 MR 膀胱造影的运动校正
DOI：
10.1109/tbme.2013.2257769
复制
发表时间：
2013-09
期刊：
影响因子：
--
通讯作者：
He X
中科院分区：
文献类型：
--
作者：
Lin Q;Liang Z;Duan C;Ma J;Li H;Roque C;Yang J;Zhang G;Lu H;He X
磁共振（MR）膀胱造影或基于MR的虚拟膀胱镜检查是一种有前途的新技术，以评估整个膀胱在一个完全无创的方式。这就要求获得的解剖膀胱图像具有较高的空间分辨率和足够的信噪比。这通常会导致长时间扫描（>5分钟），并由于膀胱无意识运动和变形而导致图像伪影。在本文中，我们研究了一种图像处理方法，以减轻运动和变形的问题。代替传统的单次长时间扫描，为了将膀胱运动从内扫描转移到内扫描，采集了六次重复的短时间扫描（每次约1分钟）。然后，通过将短时间扫描配准到选定的参考并最终形成单个平均运动校正图像来解决扫描间运动。为了评估所提出的方法，生成三种类型的图像：1）通过短时间扫描的配准和平均的运动校正图像; 2）短时间扫描的直接平均图像（没有运动校正）;以及3）对应的长时间扫描的单个图像。6名专家对这些图像进行盲法评分，评分标准包括两个重要方面：1）膀胱壁的清晰度; 2）图像质量的总体表现。对分数的统计分析表明，在这两个方面的最佳结果是通过所提出的运动校正平均值来实现的。此外，通过线性混合效应模型的测量，运动校正平均值相对于其他两个平均值的优越性具有统计学显著性，p值<0.05。我们的研究结果可能有助于通过减轻运动挑战来检测MR膀胱造影中的膀胱异常。这种方法的有效性取决于所获得的短时扫描的噪声水平和图像配准的鲁棒性，并且需要在这两个方面进行进一步的努力。
Magnetic resonance (MR) cystography or MR-based virtual cystoscopy is a promising new technology to evaluate the entire bladder in a fully noninvasive manner. It requires the anatomical bladder images be acquired at high spatial resolution and with adequate signal-to-noise ratio (SNR). This often leads to a long-time scan (>5 min) and results in image artifacts due to involuntary bladder motion and deformation. In this paper, we investigated an image-processing approach to mitigate the problem of motion and deformation. Instead of a traditional single long-time scan, six repeated short-time scans (each of approximately 1 min) were acquired for the purpose of shifting bladder motion from intrascan into interscans. Then, the interscan motions were addressed by registering the short-time scans to a selected reference and finally forming a single average motion-corrected image. To evaluate the presented approach, three types of images were generated: 1) the motion-corrected image by registration and average of the short-time scans; 2) the directly averaged image of the short-time scans (without motion correction); and 3) the single image of the corresponding long-time scan. Six experts were asked to blindly score these images in terms of two important aspects: 1) the definition of the bladder wall and 2) the overall expression on the image quality. Statistical analysis on the scores suggested that the best result in both the aspects is achieved by the presented motion-corrected average. Furthermore, the superiority of the motion-corrected average over the other two is statistically significant by the measure of a linear mixed-effect model with p-values <; 0.05. Our findings may facilitate the detection of bladder abnormality in MR cystography by mitigating the motion challenge. The effectiveness of this approach depends on the noise level of acquired short-time scans and the robustness of image registration, and future effort on these two aspects is needed.