Introduction

The human wrist is composed of an intricate arrangement of interdependent small structures. Among these are ligaments, cartilage and the triangular fibrocartilage complex (TFCC). Reliable evaluation of these structures are of great importance for diagnosis and clinical decision making, but is often difficult and can require invasive arthroscopy. Due to the complex anatomy and small size of these wrist structures, diagnostics may greatly benefit from the high-resolution imaging possible at 7T thanks to increased SNR. However, with increasing resolution the scan duration increases rapidly, decreasing patient comfort and increasing the risk of motion artifacts. Scan acceleration is therefore highly desired. A clinically relevant MRI acquisition for diagnosing wrist injuries at all field strengths is a 3D proton density-weighted TSE with SPAIR fat suppression¹. Here, we evaluated compressed sensing² acceleration of a clinical wrist 7T MRI protocol with and without fat-suppression in comparison to standard SENSE-acceleration.

Methods

All MRI scans were performed with a 7T Philips Achieva scanner (Philips, Best, The Netherlands) using a 1Tx/16Rx wrist coil (RAPID Biomedical GmbH, Rimpar, Germany). Thirteen healthy volunteers were included in the study after providing written informed consent. The clinical 7T 3D PD-w TSE protocol used as reference had isotropic spatial resolution of 0.45mm and was accelerated with SENSE-factors 2.2x2.0, giving a scan duration of 7:07 minutes (‘SENSE0.45’). To compare the SENSE0.45 scan, the scan protocol included two compressed sensing (CS) scans, performed using the in-house developed PROUD technique allowing sparse acquisitions^3,4. Compared to the reference SENSE0.45 scan, the first compressed sensing scan (‘CS0.45’) had equal resolution but additional acceleration, leading to a scan duration of 4:05 (43% scan time reduction), and the second (‘CS0.35’) had equal scan duration but an increased resolution of 0.35mm isotropic. To keep the TSE shot duration constant, the TSE factor was decreased in CS0.35 with respect to CS0.45 and SENSE0.45. For all scans, versions with and without SPAIR fat suppression were performed. The undersampling patterns used in the CS scans and additional scan parameters are shown in Figure 1. Both undersampling patterns contain a 25x25 sized fully sampled center square which was used to estimate receive coil sensitivity maps⁵ using the Berkeley Advanced Reconstruction Toolbox (BART)⁶. All CS reconstructions were performed using the PICS algorithm from BART and MRecon (Gyrotools, Zurich, Switzerland) in MATLAB (R2016a, The MathWorks Inc.). Regularization parameter optimization was performed heuristically. For the SENSE-scans, the vendor online reconstruction was used. Finally, a variation of SENSE0.45 was performed using a higher SENSE-acceleration of 3.0x2.6 (‘High-SENSE’). This had the same scan duration as CS0.45, 4:05 minutes. To assess image quality quantitatively, a radiologist performed blinded scoring of visibility of twelve wrist structures for the SPAIR scans: triangular fibrocartilage complex (TFCC, radial, ulnar and foveal attachment), scapholunate and lunotriquetral ligament (SLL/LTL, dorsal and palmar portion), bone trabeculae, articular cartilage, extensor carpi ulnaris (ECU) tendon and median and ulnar nerve. Additionally, images were scored on SNR, edge sharpness, perceived tissue contrast and artifact absence. In scoring, a 1 to 5 Likert scale was used.

Results

Figure 2 shows SPAIR images of one volunteer. The high quality in the SENSE0.45 scan was consistent over individuals, as confirmed by the average scores in Figure 3. The CS0.45 scan shows some blurring compared to SENSE0.45, but structures are still well visible and CS0.45 scores adequate to good (3-4) for all parameters except bone trabeculae. Although CS0.35 has a higher isotropic resolution in preserved scan time, this did not translate into consistently improved visibility compared to SENSE0.45 and CS0.45. Figure 4 shows non-SPAIR images for all scan types. The High-SENSE scan shows considerable SENSE-artifacts.

Discussion

Good to excellent visibility of nearly all investigated wrist structures could be obtained using a SENSE-factor of 2.2x2.0 in a 0.45mm isotropic 3D PDw SPAIR TSE scan. The CS0.45 scan reduced the scan time with 43% and retained an adequate to good image quality. Future studies will evaluate if diagnostic accuracy can be preserved using CS0.45 instead of SENSE0.45. For the SPAIR scans, CS0.35 did not outperform SENSE0.45. Visibility of bone trabeculae was scored lower for CS SPAIR scans. However, the non-SPAIR scans might be more suited for evaluation of bone structure. It is also clear that using higher SENSE-factors to obtain the same scan time reduction as with CS0.45 is not possible due to severe SENSE-artifacts. Future work will be focused on further improvement of the moderate image quality in CS-scans, with candidates of causing this being motion, a difference between vendor online and MRecon offline reconstruction and a difference in k-space trajectory between SENSE and CS.

Conclusion

Compressed sensing was successfully applied to high-resolution 3D TSE SPAIR PD-weighted MRI at 7T to further shorten SENSE-accelerated scans by 43% while still obtaining adequate to good wrist structure visibility.

Acknowledgements

The National 7T facility at Lund University Bioimaging Center is acknowledged for providing experimental resources.