Purpose

Oblique sagittal high resolution MRI sequences, orthogonal to the longitudinal axis of the brachial plexus, can reliably depict morphologic and signal abnormalities. However, this acquisition technique can be challenging due to respiratory-induced ghosting artifact that may obscure nerve detail. More than two decades ago, Posniak et al. mentioned the possible use of respiratory triggering (RT) for plexus MRI¹. However, the efficacy of RT for brachial plexus imaging has not yet been reported. Our hypothesis was that RT would minimize ghosting artifact and improve nerve conspicuity of the brachial plexus, without prohibitive scan time penalty.

Methods

In this prospective, IRB-approved study of 5 healthy volunteers and 20 patients, proton density (PD) or T2-weighted Dixon fat suppressed (DFS) sequences with and without RT (with the same scan coverage and comparable spatial resolution – Table 1), were acquired consecutively within each examination. The study was performed on three 3T 60cm bore scanners (Discovery MR750, GE Healthcare, USA) using 16 channel small and/or medium flexible coils. RT was achieved prospectively with a thoracic bellow, 1-2 respiratory rate (RR) intervals, trigger window = 40%, and trigger point = 30%. Two musculoskeletal radiologists, blinded to the presence of RT, scored the sequences independently. To optimize blinding, the non-RT and RT acquisitions of different subjects were submitted for scoring in a random fashion. Subjective scoring criteria for A) Ghosting artifact; B) Nerve conspicuity (defined as delineation of nerve fascicular architecture) for three regions of the brachial plexus (supraclavicular, retroclavicular and infraclavicular); and C) Overall diagnostic confidence were employed (Table 2). The total image quality score was computed by summing the scores of the three regions of B. RR, total scan time and number of slices were recorded, with average scan time per slice calculated. Wilcoxon signed-rank test was used to compare non-RT and RT parameters. Inter-observer agreement was assessed using weighted Cohen’s kappa analysis with the exception of total image quality score, which were evaluated with Krippendorf’s alpha analysis. Spearman correlation evaluated the associations of image quality with scan time and RR, where good image quality defined by nerve conspicuity score ≥3 (Scale 1- 4, Table 2). Significance was set a priori to p<0.05.

Results

In 25 subjects, 50 sequences were compared (4 PD, 12 3-pt. DFS, 34 2-pt. DFS). Figure 1 shows the subjective scoring comparison results. RT demonstrated improvements in all image quality catergories with statistical significance (p<0.05). The diagnostic confidence increased for both readers from a median score of 2 to 3 ( p = 0.02 for reader 1 and p = 0.04 for reader 2). Inter-observer agreement for asessment of ghosting artifact showed substantial to almost perfect agreement (kappa coefficents: 0.81 for non-RT and 0.77 for RT). Evaluation of nerve conspicuity demonstrated almost perfect agreement (Krippendorff's alpha coefficients: 0.986 for non-RT and 0.985 for RT). Inter-observer agreement of the remaining scores demonstrated moderate to substantial agreement (kappa coefficient ranging from: 0.41-0.79). There was no evidence of association between scan time with good image quality on non-RT or RT sequences (Spearman’s ρ ranging from: -0.35 to 0.29 and -0.04 to 0.29 respectively). There was a consistent weakly positive correlation between higher respiratory rate and good image quality (Spearman’s ρ ranging from: 0.28-0.43). Median scan time per slice increased from 6.95 sec (non-RT) to 10.21 sec (RT) (46% increase) and total median scan time increased from 4:44 min. (non-RT) to 6:00 min. (RT) (36% increase).

Discussion

This study demonstrated that for brachial plexus MRI, RT can reduce ghosting artifact, improve nerve fascicular architecture delineation and increase overall diagnostic confidence, with acceptable scan time cost and very good inter-observer agreement. There was a consistent weak association between higher respiratory rate and better image quality. A lower respiratory rate was challenging for RT scans as all slices in a respiratory cycle are acquired in a single TR, thereby contributing to a relatively long non-data filling period. An additional challenge we subjectively observed with respiratory triggering was an irregular respiratory rhythm, which occasionally precipitated imperfect triggering. RT techniques that better incorporate RR and respiratory rhythm may further minimize scan time and improve image quality.

Conclusion

Given the acceptable scan time impact observed in this study, use of RT could be considered for high resolution brachial plexus MRI to minimize artifact and improve image quality. Alternative techniques, such as soft gating^3,4 or guided breathing that are insensitive to respiratory patterns may be further investigated to mitigate artifact.

Acknowledgements

Hospital of Special Surgery has an institutional agreement with GE Healthcare