Just how much data need to be collected for reliable bootstrap DT-MRI?
Just how much data need to be collected for reliable bootstrap DT-MRI?
复制标题
DOI：
10.1002/mrm.21014
复制
发表时间：
2006-10-01
影响因子：
3.3
通讯作者：
Jones, Derek K.
中科院分区：
文献类型：
--
作者：
O'Gorman, Ruth L.;Jones, Derek K.
扩散张量MRI （DT-MRI）可以根据水分子扩散率的方向依赖性提供纤维取向的估计，从而可以使用纤维束成像方法重建白质纤维通路。然而，来自各种来源的噪声可能会在扩散张量元素的估计中引入不确定性，从而导致纤维取向估计的误差，从而导致纤维路径的束状图重建可能不精确和不准确。最近，人们尝试用自举法来表征dt - mri衍生参数的不确定性；然而，关于精确重建DT-MRI参数的概率分布所需的重复测量和引导带的数量，仍然存在几个问题。本研究探讨了自举法表征DT-MRI参数分布的准确性和精密度。使用已知每个参数的真实可变性的理想系统，考虑了许多包含不同数量的各向同性分布梯度向量的实验自举设计和采样方案。本研究表明，对于大多数DT-MRI实验，如果最小自举次数约为500，并且每个扩散加权强度至少使用5个重复样本进行自举，则可以获得稳健的结果。
Diffusion tensor MRI (DT-MRI) can provide estimates of fiber orientation derived from the orientational dependence of the diffusivity of water molecules, enabling the reconstruction of white matter fiber pathways using tractography methods. However, noise arising from various sources can introduce uncertainty into the estimates of the elements of the diffusion tensor, resulting in errors in fiber orientation estimates such that tractography reconstructions of fiber pathways potentially can be imprecise and inaccurate. Recently, attempts have been made to characterize the uncertainty in DT-MRI-derived parameters using the bootstrap method; however, several questions remain open regarding the number of repeat measurements and boot-straps required to accurately and precisely reconstruct the probability distributions of the DT-MRI parameters. This study investigates the accuracy and precision of the bootstrap method for characterizing distributions of DT-MRI parameters. A number of experimental bootstrap designs and sampling schemes containing different numbers of isotropically distributed gradient vectors are considered, using an idealized system where the true variability in each parameter is known. This study demonstrates that for most DT-MRI experiments, robust results will be obtained if the minimum number of bootstraps is approximately 500, and that at least five repeat samples of each diffusion-weighted intensity should be used for bootstrapping.