Translational Imaging Group, CMIC, Department of Medical Physics and Biomedical Engineering, University College London, Room 8.04 8th floor Malet Place Engineering Building, 2, Malet Place, WC1E 7JE London, UK; Dementia Research Centre, UCL Institute of Neurology, WC1N 3BG London, UK.
Santa Creu i Sant Pau Hospital, Universitat Autonòma Barcelona, 08041 Barcelona, Spain.
J Neuroradiol. 2018 Mar;45(2):114-122. doi: 10.1016/j.neurad.2017.10.001. Epub 2017 Nov 11.
Visual rating scales have limited capacities to depict the regional distribution of cerebral white matter hyperintensities (WMH). We present a regional-zonal volumetric analysis alongside a visualization tool to compare and deconstruct visual rating scales.
3D T1-weighted, T2-weighted spin-echo and FLAIR images were acquired on a 3T system, from 82 elderly participants in a population-based study. Images were automatically segmented for WMH. Lobar boundaries and distance to ventricular surface were used to define white matter regions. Regional-zonal WMH loads were displayed using bullseye plots. Four raters assessed all images applying three scales. Correlations between visual scales and regional WMH as well as inter and intra-rater variability were assessed. A multinomial ordinal regression model was used to predict scores based on regional volumes and global WMH burdens.
On average, the bullseye plot depicted a right-left symmetry in the distribution and concentration of damage in the periventricular zone, especially in frontal regions. WMH loads correlated well with the average visual rating scores (e.g. Kendall's tau [Volume, Scheltens]=0.59 CI=[0.53 0.62]). Local correlations allowed comparison of loading patterns between scales and between raters. Regional measurements had more predictive power than global WMH burden (e.g. frontal caps prediction with local features: ICC=0.67 CI=[0.53 0.77], global volume=0.50 CI=[0.32 0.65], intra-rater=0.44 CI=[0.23 0.60]).
Regional-zonal representation of WMH burden highlights similarities and differences between visual rating scales and raters. The bullseye infographic tool provides a simple visual representation of regional lesion load that can be used for rater calibration and training.
视觉评分量表在描绘脑白质高信号(WMH)的区域性分布方面能力有限。我们提出了一种区域-分区容积分析,并结合可视化工具来比较和分解视觉评分量表。
在 3T 系统上采集了 82 名基于人群的研究中老年参与者的 3D T1 加权、T2 加权自旋回波和 FLAIR 图像。使用自动分割方法对 WMH 进行分割。使用脑叶边界和距脑室表面的距离来定义白质区域。使用牛眼图显示区域性分区 WMH 负荷。四名评分者对所有图像应用三种量表进行评估。评估视觉量表与区域性 WMH 之间的相关性以及评分者之间和评分者内部的变异性。使用多项有序回归模型基于区域性体积和总体 WMH 负担预测评分。
平均而言,牛眼图描绘了脑室周围区域损伤的分布和集中在左右两侧的对称性,尤其是在前额区域。WMH 负荷与平均视觉评分(如肯德尔 tau [体积,Scheltens] = 0.59,CI=[0.53 0.62])相关性良好。局部相关性允许比较量表之间和评分者之间的负荷模式。区域测量比总体 WMH 负担具有更高的预测能力(例如,使用局部特征预测额帽:ICC=0.67,CI=[0.53 0.77],全球体积=0.50,CI=[0.32 0.65],评分者内=0.44,CI=[0.23 0.60])。
WMH 负担的区域性分区表示突出了视觉评分量表和评分者之间的相似之处和差异。牛眼图信息图工具提供了一种简单的区域性病变负荷的可视化表示,可用于评分者校准和培训。