Department of Radiology, Technical University of Munich, Ismaninger Strasse 22, 81675, Munich, Germany; Department of Radiology and Biomedical Imaging, University of California, 185 Berry Street, Suite 350, San Francisco, CA, 94107, USA.
Department of Radiology, Technical University of Munich, Ismaninger Strasse 22, 81675, Munich, Germany.
Eur J Radiol. 2019 Sep;118:277-284. doi: 10.1016/j.ejrad.2019.06.006. Epub 2019 Jun 16.
To assess the feasibility and diagnostic value of compressed sensing for accelerating two-dimensional turbo spin echo imaging of the ankle.
Ankles of 20 volunteers were scanned (mean age 30.2 ± 7.3 years, 13 men) at 3 T MRI. Coronal and sagittal intermediate-weighted (IM) sequences with fat saturation as well as axial T2- and coronal T1-weighted sequences were acquired using parallel imaging based on sensitivity encoding (SENSE) only as well as with a combination of compressed sensing (CS) and SENSE. Compressed sensing is a technique that acquires less data through k-space random undersampling and enables a reduction in total acquisition time by 20%. All images were reviewed by two radiologists, image quality was graded using a 5-point Likert scale and signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of different anatomical structures of the ankle were assessed and compared between sequences with SENSE only and with the combination of CS and SENSE using Wilcoxon signed-rank tests and Cohen's kappa.
There was a substantial to perfect agreement for the rating between the images acquired with SENSE only and with the combination of CS and SENSE when assessing cartilage, subchondral bone and ligaments (κ = 0.75 - 0.89). SNR was slightly higher for the combination of CS and SENSE sequences compared to the sequences acquired with SENSE only, yet this finding was not significant (P = 0.18-0.62). Moreover, CNR of cartilage/fluid, subchondral bone/cartilage, ligaments/fluid and ligaments/fat did not show significant differences between the sequences acquired with SENSE only and the combination of CS and SENSE (P > 0.05). The interreader agreement was substantial to excellent for both techniques (κ=0.75 - 0.89).
Compressed sensing reduced the acquisition time of conventional MR imaging of the ankle by 20% without decreasing diagnostic image quality, SNR and CNR.
评估压缩感知在加速二维涡轮自旋回波成像踝关节中的可行性和诊断价值。
在 3T MRI 上对 20 名志愿者（平均年龄 30.2±7.3 岁，13 名男性）的踝关节进行扫描。采集冠状位和矢状位中等加权（IM）序列，均带有脂肪饱和，以及轴向 T2 加权和冠状位 T1 加权序列，采用基于敏感度编码（SENSE）的并行成像，以及压缩感知（CS）和 SENSE 的组合。压缩感知是一种通过 k 空间随机欠采样获取较少数据的技术，通过减少总采集时间 20%来实现。所有图像均由两位放射科医生进行评估，使用 5 分 Likert 量表对图像质量进行分级，并评估和比较仅使用 SENSE 和 CS 与 SENSE 组合采集的不同解剖结构的信噪比（SNR）和对比噪声比（CNR），使用 Wilcoxon 符号秩检验和 Cohen's kappa 进行比较。
仅使用 SENSE 采集的图像和使用 CS 与 SENSE 组合采集的图像在评估软骨、软骨下骨和韧带时具有高度至极好的一致性（κ=0.75-0.89）。与仅使用 SENSE 采集的序列相比，CS 与 SENSE 组合序列的 SNR 略高，但无统计学意义（P=0.18-0.62）。此外，软骨/液体、软骨下骨/软骨、韧带/液体和韧带/脂肪的 CNR 在仅使用 SENSE 采集的序列和 CS 与 SENSE 组合采集的序列之间无显著差异（P>0.05）。两种技术的观察者间一致性均为高度至极好（κ=0.75-0.89）。
压缩感知将常规踝关节磁共振成像的采集时间缩短了 20%，而不会降低诊断图像质量、SNR 和 CNR。
