From the School of Medicine (S.Y., E.M.M.), Department of Cardiovascular Medicine (P.R.T.), and Department of Radiology (J.L., A.C.R., A.H.), University of California, San Diego, 9300 Campus Point Dr, La Jolla, CA 92037-0841; and Department of Radiology, Stanford University School of Medicine, Stanford, Calif (M.T.A., S.S.V.).
Radiology. 2022 Mar;302(3):584-592. doi: 10.1148/radiol.2021211270. Epub 2021 Nov 30.
Background Four-dimensional (4D) flow MRI has the potential to provide hemodynamic insights for a variety of abdominopelvic vascular diseases, but its clinical utility is currently impaired by background phase error, which can be challenging to correct. Purpose To assess the feasibility of using deep learning to automatically perform image-based background phase error correction in 4D flow MRI and to compare its effectiveness relative to manual image-based correction. Materials and Methods A convenience sample of 139 abdominopelvic 4D flow MRI acquisitions performed between January 2016 and July 2020 was retrospectively collected. Manual phase error correction was performed using dedicated imaging software and served as the reference standard. After reserving 40 examinations for testing, the remaining examinations were randomly divided into training (86% [85 of 99]) and validation (14% [14 of 99]) data sets to train a multichannel three-dimensional U-Net convolutional neural network. Flow measurements were obtained for the infrarenal aorta, common iliac arteries, common iliac veins, and inferior vena cava. Statistical analyses included Pearson correlation, Bland-Altman analysis, and tests with Bonferroni correction. Results A total of 139 patients (mean age, 47 years ± 14 [standard deviation]; 108 women) were included. Inflow-outflow correlation improved after manual correction ( = 0.94, < .001) compared with that before correction ( = 0.50, < .001). Automated correction showed similar results ( = 0.91, < .001) and demonstrated very strong correlation with manual correction ( = 0.98, < .001). Both correction methods reduced inflow-outflow variance, improving mean difference from -0.14 L/min (95% limits of agreement: -1.61, 1.32) (uncorrected) to 0.05 L/min (95% limits of agreement: -0.32, 0.42) (manually corrected) and 0.05 L/min (95% limits of agreement: -0.38, 0.49) (automatically corrected). There was no significant difference in inflow-outflow variance between manual and automated correction methods ( = .10). Conclusion Deep learning automated phase error correction reduced inflow-outflow bias and variance of volumetric flow measurements in four-dimensional flow MRI, achieving results comparable with manual image-based phase error correction. © RSNA, 2021 See also the editorial by Roldán-Alzate and Grist in this issue.
背景 四维（4D）流动 MRI 有可能为各种腹盆腔血管疾病提供血流动力学见解，但目前其临床应用受到背景相位误差的限制，这种误差很难纠正。目的 评估使用深度学习自动执行 4D 流动 MRI 基于图像的背景相位误差校正的可行性，并比较其相对于手动图像校正的效果。材料与方法 回顾性收集了 2016 年 1 月至 2020 年 7 月期间进行的 139 例腹盆腔 4D 流动 MRI 采集的便利样本。使用专用成像软件进行手动相位误差校正，作为参考标准。在保留 40 项检查用于测试后，其余检查随机分为训练（86%[99 项中的 85 项]）和验证（14%[99 项中的 14 项]）数据集，以训练多通道三维 U-Net 卷积神经网络。获得肾下主动脉、髂总动脉、髂总静脉和下腔静脉的流量测量值。统计分析包括 Pearson 相关性、Bland-Altman 分析和带有 Bonferroni 校正的 t 检验。结果 共纳入 139 例患者（平均年龄，47 岁±14[标准差]；108 例女性）。与校正前（ = 0.50，<.001）相比，手动校正后流入-流出相关性提高（ = 0.94，<.001）。自动校正也显示出类似的结果（ = 0.91，<.001），并与手动校正高度相关（ = 0.98，<.001）。两种校正方法均降低了流入-流出方差，将平均差值从-0.14 L/min（95%置信区间：-1.61，1.32）（未校正）改善至 0.05 L/min（95%置信区间：-0.32，0.42）（手动校正）和 0.05 L/min（95%置信区间：-0.38，0.49）（自动校正）。手动和自动校正方法之间的流入-流出方差无显著差异（ =.10）。结论 深度学习自动相位误差校正可减少 4D 流动 MRI 中容积流量测量的流入-流出偏倚和方差，结果可与手动基于图像的相位误差校正相媲美。 ©2021RSNA，见本期 Roldán-Alzate 和 Grist 的社论。
