From the Department of Diagnostic and Interventional Radiology, Lausanne University Hospital and University of Lausanne, Rue du Bugnon 46, 1011 Lausanne, Switzerland (M.R., T.H., M.H., F.B., T.K., P.O.); Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland (T.H., T.K.); and LTS5, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland (T.H., T.K.).
Radiology. 2019 Dec;293(3):620-630. doi: 10.1148/radiol.2019182843. Epub 2019 Oct 1.
Background Use of a T2 mapping sequence in addition to the conventional knee MRI protocol increases sensitivity to early cartilage lesions but is time consuming. Purpose To test the in vitro validity of quantitative data from an accelerated parallel T2 mapping sequence (combined generalized autocalibrating partially parallel acquisition and model-based accelerated relaxometry by iterative nonlinear inversion [GRAPPATINI]) of the knee and to compare in vivo synthetic images generated with this sequence with those generated with conventional morphologic sequences. Materials and Methods T2 estimations with GRAPPATINI were validated in vitro in comparison with T2 estimations with routine multisection multiecho and reference standard single-section single-echo spin-echo T2 mapping sequences by using a Bland-Altman plot. Synthetic morphologic images (intermediate-weighted sequence, 34-msec echo time; T2-weighted sequence, 80-msec echo time) were compared in vivo with corresponding conventional morphologic turbo spin-echo 3-T sequences by three readers in consecutive patients recruited retrospectively from February to May 2018. Synthetic and conventional morphologic images were compared by using rates of interreader agreement, κ statistics, and rates of findings. Results T2 values with GRAPPATINI were accurate compared with those obtained with the reference single-section single-echo sequence, with slight T2 overestimation (2.7 msec). Sixty-one patients (mean age, 43 years ± 16 [standard deviation]; 32 men) were included. The rate of agreement when one reader used synthetic morphologic images and the other used conventional sequences was not inferior to the rate of agreement when all readers used conventional sequences (upper bounds of 95% confidence intervals of differences between rates of agreement ≤ 4.8%). Interreader agreement was similar for the conventional set alone, the synthetic set alone, and when readers used different sets (two-by-two differences between κ values for all items ≤ 0.15). The rates of findings were not different between synthetic and conventional image sets (all ≥ .07) except for two items (femoral trochlear cartilage [3.0% vs 0.3%, = .006] and joint effusion [0.3% vs 2.7%, = .005]). Conclusion This T2 mapping sequence yields, in one acquisition, accurate T2 values and synthetic morphologic images that are comparable with those obtained with conventional turbo spin-echo sequences. © RSNA, 2019 See also the editorial by Fritz in this issue.
背景 在常规膝关节 MRI 方案的基础上使用 T2 映射序列可以提高对早期软骨病变的敏感性，但耗时。目的 测试膝关节加速并行 T2 映射序列（组合广义自校准部分并行采集和基于模型的迭代非线性反转弛豫率加速[GRAPPATINI]）定量数据的体外有效性，并比较该序列生成的合成图像与常规形态序列生成的图像。材料与方法 通过 Bland-Altman 图，将 GRAPPATINI 的 T2 估计值与多节多回波常规 T2 映射序列和参考标准单节单回波自旋回波 T2 映射序列的 T2 估计值进行了体外验证。2018 年 2 月至 5 月，连续回顾性招募患者，由三位读者比较体内合成形态图像（中等权重序列，34 毫秒回波时间；T2 加权序列，80 毫秒回波时间）与相应的常规形态 turbo 自旋回波 3-T 序列。通过三位读者的读者间协议率、κ 统计量和检出率来比较合成和常规形态图像。结果 GRAPPATINI 的 T2 值与参考单节单回波序列的 T2 值相比准确，仅存在轻微的 T2 值高估（2.7 毫秒）。61 例患者（平均年龄，43 岁±16[标准差]；32 例男性）被纳入研究。一位读者使用合成形态图像，另一位读者使用常规序列时的协议率不低于所有读者均使用常规序列时的协议率（差异的 95%置信区间上限值≤4.8%）。仅使用常规序列组、仅使用合成序列组和读者使用不同序列组时的读者间协议率相似（所有项目的κ 值的两两差异均≤0.15）。合成图像组和常规图像组的检出率无差异（均≥0.07），除了两个项目（股骨滑车软骨[3.0%比 0.3%， =.006]和关节积液[0.3%比 2.7%， =.005]）。结论 该 T2 映射序列在一次采集时可获得准确的 T2 值和合成形态图像，其与常规 turbo 自旋回波序列的结果相当。
