商用PDFF MRI脂肪定量的长期跨平台重复性、偏倚与线性:多中心多厂商体模研究
该研究在35个月内对多中心观察性试验中的438名受试者和16名志愿者使用3T或1.5T MR系统进行PDFF MRI测量,并同步扫描脂肪-水体模。结果显示所有厂商和场强下线性良好,但存在显著的厂商和场强依赖性偏倚,汇总重复性系数为6.93%;在不同MR系统间进行纵向PDFF测量时,仅变化≥7%可视为真实差异。
要点
- 不同厂商或场强MR系统的PDFF测量存在显著偏倚差异,纵向随访换机时需谨慎解读。
- 跨系统纵向比较时,PDFF变化需≥7%(重复性系数6.93%)才考虑为真实差异。
- PDFF在0-100%脂肪范围内线性良好(斜率0.94),非线性效应可忽略。
局限性
- 原文未述局限性
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
OBJECTIVES: Proton density fat fraction (PDFF) is a validated biomarker of tissue fat quantification. However, validation has been limited to single-center or multi-center series using non-FDA-approved software. Thus, we assess the bias, linearity, and long-term reproducibility of PDFF obtained using commercial PDFF packages from several vendors.
METHODS: Over 35 months, 438 subjects and 16 volunteers from a multi-center observational trial underwent PDFF MRI measurements using a 3-T MR system from one of three different vendors or a 1.5-T system from one vendor. Fat-water phantom sets were measured as part of each subject's examination. Manual region-of-interest measurements on the %fat image, then cross-sectional bias, linearity, and long-term reproducibility were assessed.
RESULTS: Three hundred ninety-two phantom measurements were evaluable (90%). Bias ranged from 2.4 to - 3.8% for the lowest to the highest weight %fat. Regression fits of PDFF against synthesis weight %fat showed negligible non-linear effects and a linear slope of 0.94 (95% confidence interval: 0.938, 0.947). We observed significant vendor (p < 0.001) and field strength (p < 0.001) differences in bias and longitudinal variability. When the results were pooled across sites, vendors, and field strengths, the estimated reproducibility coefficient was 6.93% (95% CI: 6.25%, 7.81%).
CONCLUSIONS: This study demonstrated good linearity, accuracy, and reproducibility for all investigated manufacturers and field strengths. However, significant vendor-dependent and field strength-dependent bias were found. While longitudinal PDFF measurements may be made using different field strength or vendor MR systems, if the MR system is not the same, based on these results, only PDFF changes ≥ 7% can be considered a true difference.
KEY POINTS: • Phantom fat fraction (PDFF) MRI measurements over 35 months demonstrated good linearity, accuracy, and reproducibility for the vendor systems investigated. • Non-linear effects were negligible (linear slope of 0.94) over 0-100% fat; however, significant vendor (p < 0.001) and field strength (p<0.001) differences in bias and longitudinal variability were identified. Bias ranged from 2.4 to - 3.8% for 0-100 weight% fat, respectively. • Measurement bias could affect the accuracy of PDFF in clinical use. As the reproducibility coefficient was 6.93%, only greater changes in % fat can be considered true differences when making longitudinal PDFF measurements on different MR systems.
引用
- Schneider E, Remer EM, Obuchowski NA, et al. Long-term inter-platform reproducibility, bias, and linearity of commercial PDFF MRI methods for fat quantification: a multi-center, multi-vendor phantom study. Eur Radiol. 2021. PMID: 33768291. doi:10.1007/s00330-021-07851-8
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。