Delles Michael, Rengier Fabian, Ley Sebastian, von Tengg-Kobligk Hendrik, Kauczor Hans-Ulrich, Dillmann Rüdiger, Unterhinninghofen Roland
Department of Diagnostic and Interventional Radiology, University Hospital Heidelberg, 69120 Heidelberg, Germany.
Annu Int Conf IEEE Eng Med Biol Soc. 2011;2011:6829-32. doi: 10.1109/IEMBS.2011.6091684.
In cardiovascular diagnostics, phase-contrast MRI is a valuable technique for measuring blood flow velocities and computing blood pressure values. Unfortunately, both velocity and pressure data typically suffer from the strong image noise of velocity-encoded MRI. In the past, separate approaches of regularization with physical a-priori knowledge and data representation with continuous functions have been proposed to overcome these drawbacks. In this article, we investigate polynomial regularization as an exemplary specification of combining these two techniques. We perform time-resolved three-dimensional velocity measurements and pressure gradient computations on MRI acquisitions of steady flow in a physical phantom. Results based on the higher quality temporal mean data are used as a reference. Thereby, we investigate the performance of our approach of polynomial regularization, which reduces the root mean squared errors to the reference data by 45% for velocities and 60% for pressure gradients.
在心血管诊断中，相位对比磁共振成像（MRI）是一种用于测量血流速度和计算血压值的重要技术。不幸的是，速度和压力数据通常都受到速度编码MRI的强烈图像噪声的影响。过去，人们提出了利用物理先验知识进行正则化和用连续函数进行数据表示的单独方法来克服这些缺点。在本文中，我们研究多项式正则化作为结合这两种技术的一个示例规范。我们对物理模型中稳定流的MRI采集进行时间分辨三维速度测量和压力梯度计算。基于更高质量的时间平均数据的结果用作参考。由此，我们研究了我们的多项式正则化方法的性能，该方法将相对于参考数据的速度均方根误差降低了45%，压力梯度降低了60%。
