Efficient and complete fat suppression in two-point Dixon MR imaging: sequence considerations and assessment in lumbar plexus imaging.

Abstract

Two-point Dixon MRI is widely used for lumbar plexus imaging but often yields incomplete fat suppression due to neglecting fat's multi-peak spectrum. This study developed and evaluated a post-processing algorithm that accounts for the multi-peak fat spectrum to achieve more effective fat suppression. A novel Dark-fat processing algorithm was developed to enhance fat suppression in water-only MR images from standard two-point Dixon reconstruction. It was evaluated using a custom fat/water phantom, prospectively in a healthy volunteer, and retrospectively in 30 patients who underwent clinical lumbar plexus imaging between March and October 2023. Imaging was done using FFE (or GRE)-Dixon, TSE (or FSE)-Dixon, and mDixon-Quant. Quantitative metrics (SNR, apparent SNR (aSNR), and apparent CNR (aCNR)) and qualitative assessments by two radiologists were analyzed. Statistical significance was determined using Welch's t-test (p < 0.05) and Wilcoxon signed-rank test with Benjamini-Krieger-Yekutieli correction (q < 0.05). The Dark-fat processing (i) substantially decreased fat signal in water-only images (84% reduction) in phantom from two-point TSE-Dixon and was statistically not different from the reference standard, multi-echo mDixon-Quant (p = 0.58); (ii) improved overall fat suppression (q < 0.05, effect size = 0.67 and 0.92), while there was no significant difference in the anatomical visualization; (iii) significantly reduced aSNR in the subcutaneous fat (p < 0.0001, effect size > 1) and muscle (p < 0.01, effect size = 0.81); and (iv) did not alter aCNR between nerve and muscle tissue (p = 0.27, effect size = 0.29). Dark-fat processing enables efficient and complete fat suppression in two-point Dixon acquisitions, especially TSE-Dixon, significantly enhancing lumbar plexus image quality. When implemented offline and as a retrospective method, it requires no change to scan time and can be applied to any two-point Dixon dataset.