Improved Brachial Plexus and C-Spine DTI using Deep Learning-based Phase Correction

Motivation: Diffusion tensor imaging (DTI) provides valuable functional information for diagnosing various pathologies affecting the brachial plexus and c-spine. However, it is SNR-starved due to the diffusion encoding gradients and long TE. Goal(s): To evaluate a DL-based phase correction (DLPC) method that reduces the noise floor by enabling robust complex averaging as well as improves image sharpness. Approach: Volunteer brachial plexus and c-spine data were acquired with DTI sequence and reconstruction was performed offline. Results: The DTI data processed with DLPC resulted in lower noise floor and sharper images, which led to improved postprocessing such as cleaner fractional anisotropy map and fiber tractography. Impact: When combined with MR neurography, DTI of the brachial plexus and c-spine with reduced noise and increased sharpness can help radiologists with differential diagnosis of pathologies resulting from nerve damage.