Magnetic resonance tractography of the brachial plexus: step-by-step

BACKGROUND: Magnetic resonance (MR) tractography of the brachial plexus (BP) is challenging due to different factors such as motion artifacts, pulsation artifacts, signal-to-noise ratio, spatial resolution; eddy currents induced geometric distortions, sequence parameters and choice of used coils. Notably challenging is the separation of the peripheral nerve bundles and skeletal muscles as both structures have similar fractional anisotropy values. We proposed an algorithm for robust visualization and assessment of BP root bundles using the segmentation of the spinal cord (SSC, C4-T1) as seed points for the initial starting area for the fibre tracking algorithm. METHODS: Twenty-seven healthy volunteers and four patients with root avulsions underwent magnetic resonance imaging (MRI) examinations on a 3T MR scanner with optimized measurement protocols for diffusion-weighted images and coronal T2 weighted 3D short-term inversion recovery sampling perfection with application optimized contrast using varying flip angle evaluation sequences used for BP fibre reconstruction and MR neurography (MRN). The fibre bundles reconstruction was optimized in terms of eliminating the skeletal muscle fibres contamination using the SSC and the tracking threshold of the normalized quantitative anisotropy (NQA) on reconstruction of the BP. In our study, the NQA parameter has been used for fiber tracking instead of fractional anisotropy (FA). The diffusion data were processed in individual C4-T1 root bundles using the generalized q-sampling imaging (GQI) algorithm. Calculated diffusion parameters were statistically analysed using the two-sample t-test. The MRN was performed in MedINRIA and post-processed using the maximum intensity projection (MIP) method to demonstrate BP root bundles in multiple planes. RESULTS: In control subjects, no significant effect of laterality in diffusion parameters was found (P>0.05) in the BP. In the central part of the BP, a significant difference between control subjects and patients at P=0.02 was found in the NQA values. Other diffusion parameters were not significantly different. CONCLUSIONS: Using NQA instead of FA in the proposed algorithm allowed for a better separation of muscle and root nerve bundles. The presented algorithm yields a high quality reconstruction of the BP bundles that may be helpful both in research and clinical practice.