Accelerated MRI of the knee. Quality and efficiency of compressed sensing

To evaluate potential clinical acceleration factors of Compressed SENSE (CS)¹ in direct comparison with SENSE for fat saturated (fs)², proton density-weighted (PD)³ 2D and 3D sequences of the knee.

Twenty healthy volunteers were scanned with a 3 T scanner, all receiving a standard, fs 2D PD, three CS (CS 2, CS 3, CS 5) as well as time-equivalent SENSE accelerations (S 2, S 3, S 5). The fs 3D PD sequence was acquired with four CS (CS 6, CS 8, CS 10, CS 15) and equivalent SENSE (S 5.72, S 7.69, S 9.57, S 14) factors. Three independent readers rated the images. Signal-to-noise, contrast-to-noise, root-mean-square error and structural similarity index were analyzed for objective evaluation.

Scan time decreased with increasing CS factor (2D CS 2: 145 s, 2D CS 3: 95 s, 2D CS 5: 57 s, 3D CS 6: 293 s, 3D CS 8: 220 s, 3D CS 10: 176 s, 3D CS 15: 119 s). The 2D standard sequence was rated best for diagnostic certainty and overall image impression with an average of 4.97 ± 0.10 and 4.80 ± 0.24 (all p < 0.05), except for 2D CS 2 and 2D S 2. For the 3D sequences, the standard sequence performed better for both parameters for CS 15, S 9.57 and S 4, as well as S 7.69 for overall image impression while CS 8 was non-inferior for all tested criteria and CS 10 only inferior for delineation of the anterior cruciate ligament, both outperforming the time-equivalent SENSE accelerations.

Compressed SENSE can significantly decrease (34.39 % for 2D CS 2 and 54.17 % for 3D CS 10) scan time in knee imaging with unchanged diagnostic certainty and overall image impression compared to the clinical reference.