Purpose: To clinically evaluate MRI of the knee using a highly resolved isotropic fat-saturated (fs) proton-density weighted 3D-TSE-sequence (SPACE) at 3T.
Materials and methods: Imaging was performed on a 3T-scanner (Magnetom TRIO). For technical evaluation, sagittally orientated SPACE-datasets (repetition-time [TR], 1200 milliseconds/[TE], 30 milliseconds/voxel-size, 0.5 mm3/acquisition time, 10:35 minutes) were acquired from the dominant knee of 10 healthy volunteers. In the 3 major anatomic planes, 0.5, 1, and 2 mm thick reconstructions were performed. Signal-to-noise (SNR), SNR-efficiency, contrast-to-noise (CNR) ratios, and anatomic detail visualization were compared with a state-of-the-art 2D-TSE-sequence in 3 imaging planes (TR, 3200 milliseconds/TE, 30 milliseconds/acquisition time, 12:34 minutes). Sixty patients with cartilage and meniscus pathologies were examined with these techniques. Patient SPACE-datasets were assessed in 1-mm thick reconstructions. Arthroscopical correlation was available for 18 patients. Lesion detection and diagnostic confidence were assessed by 2 radiologists independently. Statistical analysis was performed using 95% confidence intervals, Wilcoxon signed rank tests, and Weighted-kappa.
Results: SNR-efficiency of SPACE was 4 to 5 times higher than for 2D-TSE-sequences. SNR and CNR of 1-mm thick SPACE-reconstructions were comparable to 2D-TSE-sequences and provided superior visualization of small structures such as meniscal roots.Correlation with arthroscopy did not show significant differences between 2D- and 3D-sequences. One reader detected significantly more cartilage abnormalities with the 2D-TSE-sequence (131 vs. 151, P = 0.04), probably because of an unfamiliar fluid/cartilage contrast. Diagnostic confidence was significantly higher for meniscus abnormalities for SPACE for 1 reader. Intersequence-correlation was excellent (kappa = 0.82-0.92). Interreader-correlation was good to excellent (kappa = 0.71-0.80), intrareader-correlation was excellent (kappa = 0.90-0.92) for both sequences.
Conclusions: Time-efficient 3D-TSE-imaging of the knee at 3T is feasible with adequate SNR and CNR and excellent anatomic detail visualization. Detection and visualization of meniscus and cartilage pathologies is comparable to standard 2D-TSE-sequences. 3D-TSE-sequences with consecutive multiplanar reconstruction may become a valuable component of future knee-MRI protocols.