Improved Precision in the Measurement of Longitudinal Global and Regional Volumetric Changes via a Novel MRI Gradient Distortion Characterization and Correction Technique

Reducing measurement variability in MRI-based morphometric analysis of human brain structures will increase statistical power to detect changes between groups and longitudinally over time in individual subjects. One source of measurement error in anatomical MR is magnetic field gradientinduced geometric distortion. This work proposes a method to characterize and compensate for these distortions using a novel image processing technique relying on the image acquisition of a phantom with known geometrical dimensions, without the need to acquire the magnetic field mapping. The method is not specific to any particular shape of the phantom, as long as it provides enough coverage of the volume of interest and enough structure to densely sample the distortion field. The distortions are expressed in terms of spherical harmonic functions, which are then used to define the distortion correction field for the volume of interest. Accuracy of the distortion measurement was evaluated using numerical simulation and reproducibility was estimated using multiple scans of the phantom in the same scanner. Finally, scan-rescan experiments with nine healthy subjects demonstrated that 90% of the distortion (in terms of local volume change) can be corrected with this technique.

在基于磁共振成像（MRI）的人脑结构形态测量分析中，减少测量变异性将提高检测组间差异以及个体受试者随时间纵向变化的统计效力。解剖磁共振成像中测量误差的一个来源是磁场梯度引起的几何畸变。这项工作提出了一种方法，利用一种新颖的图像处理技术来表征和补偿这些畸变，该技术依赖于对具有已知几何尺寸的体模进行图像采集，而无需获取磁场映射。该方法不特定于体模的任何特定形状，只要它能充分覆盖感兴趣的体积，并具有足够的结构来对畸变场进行密集采样即可。畸变用球谐函数来表示，然后这些函数被用于定义感兴趣体积的畸变校正场。通过数值模拟评估了畸变测量的准确性，并通过在同一扫描仪中对体模进行多次扫描估计了可重复性。最后，对9名健康受试者进行的扫描 - 再扫描实验表明，利用这种技术可以校正90%的畸变（就局部体积变化而言）。