School of Psychology and Centre for Brain Research, The University of Auckland, Auckland, New Zealand; Brain Research New Zealand - Rangahau Roro Aotearoa, Centre of Research Excellence, New Zealand; Centre for Advanced MRI, Auckland UniServices Limited, Auckland, New Zealand.
School of Psychology and Centre for Brain Research, The University of Auckland, Auckland, New Zealand; Brain Research New Zealand - Rangahau Roro Aotearoa, Centre of Research Excellence, New Zealand.
Phys Med. 2022 Sep;101:8-17. doi: 10.1016/j.ejmp.2022.06.012. Epub 2022 Jul 16.
Individualised predictive models of cognitive decline require disease-monitoring markers that are repeatable. For wide-spread adoption, such markers also need to be reproducible at different locations. This study assessed the repeatability and reproducibility of MRI markers derived from a dementia protocol.
Six participants were scanned at three different sites with a 3T MRI scanner. The protocol employed: T1-weighted (T1w) imaging, resting state functional MRI (rsfMRI), arterial spin labelling (ASL), diffusion-weighted imaging (DWI), T2-weighted fluid attenuation inversion recovery (FLAIR), T2-weighted (T2w) imaging, and susceptibility weighted imaging (SWI). Participants were scanned repeatedly, up to six times over a maximum period of five years. One participant was also scanned a further three times on sequential days on one scanner. Fifteen derived metrics were computed from the seven different modalities.
Reproducibility (coefficient of variation; CoV, across sites) was best for T1w derived grey matter, white matter and hippocampal volume (CoV < 1.5%), compared to rsfMRI and SWI derived metrics (CoV, 19% and 21%). For a given metric, long-term repeatability (CoV across time) was comparable to reproducibility, with short-term repeatability considerably better.
Reproducibility and repeatability were assessed for a suite of markers calculated from a dementia MRI protocol. In general, structural markers were less variable than functional MRI markers. Variability over time on the same scanner was comparable to variability measured across different scanners. Overall, the results support the viability of multi-site longitudinal studies for monitoring cognitive decline.
认知能力下降的个体化预测模型需要可重复的疾病监测标志物。为了广泛应用，此类标志物还需要在不同地点具有可重复性。本研究评估了源自痴呆症方案的 MRI 标志物的可重复性和可再现性。
6 名参与者在三个不同地点使用 3T MRI 扫描仪进行扫描。该方案采用：T1 加权（T1w）成像、静息态功能磁共振成像（rsfMRI）、动脉自旋标记（ASL）、弥散加权成像（DWI）、T2 加权液体衰减反转恢复（FLAIR）、T2 加权（T2w）成像和磁化传递加权成像（SWI）。参与者重复扫描，在最长五年的时间内最多可扫描六次。一名参与者还在一台扫描仪上连续三天进行了三次扫描。从七种不同模式中计算了 15 个衍生指标。
与 rsfMRI 和 SWI 衍生指标（变异系数，19%和 21%）相比，T1w 衍生灰质、白质和海马体积的可再现性（跨站点变异系数，<1.5%）最好。对于给定的指标，长期可重复性（跨时间的变异系数）与可再现性相当，短期可重复性则好得多。
评估了从痴呆症 MRI 方案中计算得出的一系列标志物的可重复性和可再现性。通常，结构标志物的变异性小于功能磁共振成像标志物。同一扫描仪上随时间的变化与跨不同扫描仪测量的变化相当。总体而言，结果支持使用多站点纵向研究来监测认知能力下降。
