Quirk James D, Sukstanskii Alexander L, Bretthorst G Larry, Yablonskiy Dmitriy A
Mallinckrodt Institute of Radiology, Washington University School of Medicine, 4525 Scott Avenue, Campus Box 8227, St. Louis, MO 63110, USA.
J Magn Reson. 2009 May;198(1):49-56. doi: 10.1016/j.jmr.2009.01.001. Epub 2009 Jan 13.
Since their initial description, phased array coils have become increasingly popular due to their ease of customization for various applications. Numerous methods for combining data from individual channels have been proposed that attempt to optimize the SNR of the resultant images. One issue that has received comparatively little attention is how to apply these combination techniques to a series of images obtained from phased array coils that are then analyzed to produce quantitative estimates of tissue parameters. Herein, instead of the typical goal of maximizing the SNR in a single image, we are interested in maximizing the accuracy and precision of parameter estimates that are obtained from a series of such images. Our results demonstrate that a joint Bayesian analysis offers a "worry free" method for obtaining optimal parameter estimates from data generated by multiple coils (channels) from a single object (source). We also compare the properties of common channel combination techniques under different conditions to the results obtained from the joint Bayesian analysis. If the noise variance is constant for all channels, a sensitivity weighted average provides parameter estimates equivalent to the joint analysis. If both the noise variance and signal intensity are similar in all channels, a simple channel average gives an adequate result. However, if the noise variance differs between channels, an "ideal weighted" approach should be applied, where data are combined after weighting by the channel amplitude divided by the noise variance. Only this "ideal weighting" provides results similar to the automatic-weighting inherent in the joint Bayesian approach.
自从相控阵线圈首次被描述以来，由于其易于针对各种应用进行定制，它们越来越受欢迎。已经提出了许多用于组合来自各个通道的数据的方法，这些方法试图优化所得图像的信噪比。一个相对较少受到关注的问题是如何将这些组合技术应用于从相控阵线圈获得的一系列图像，然后对这些图像进行分析以产生组织参数的定量估计。在此，我们感兴趣的不是在单个图像中最大化信噪比这一典型目标，而是最大化从一系列此类图像中获得的参数估计的准确性和精度。我们的结果表明，联合贝叶斯分析提供了一种“无需担忧”的方法，用于从单个对象（源）的多个线圈（通道）生成的数据中获得最优参数估计。我们还将不同条件下常见通道组合技术的特性与联合贝叶斯分析获得的结果进行了比较。如果所有通道的噪声方差恒定，灵敏度加权平均提供的参数估计与联合分析等效。如果所有通道的噪声方差和信号强度都相似，简单的通道平均会给出足够的结果。然而，如果通道之间的噪声方差不同，则应应用“理想加权”方法，即在按通道幅度除以噪声方差进行加权后组合数据。只有这种“理想加权”提供的结果与联合贝叶斯方法中固有的自动加权相似。
