Reproducibility and dependence on diffusion weighting of line scan diffusion in the lumbar intervertebral discs

Abstract

To investigate the dependence of line scan diffusion imaging (LSDI) in the lumbar vertebral discs on diffusion weighting, fat suppression (FS), and postprocessing noise correction. Eleven normal volunteers were scanned using 4 b-value and 12 b-value LSDI protocols, with and without FS. Three repeated four b-value scans were performed for evaluation of the reproducibility of apparent diffusion coefficient (ADC) values calculated with mono- and biexponential decay models. Two-point ADC analysis for 12 b-value scans was performed with and without noise correction to evaluate the ADC dependence on diffusion weighting. Correlations between different ADC calculation and acquisition methods were evaluated. Monoexponential ADC measures had a coefficient of variation (CV) under 3%, while use of a constrained biexponential increased the CV to 6% to 9%. Strong dependence on b-value was seen from chemically shifted marrow fat signal and noise. These systematic variations in ADC were eliminated using noise correction and FS. ADC values from 4 and 12 b-value FS scans correlated strongly (R2 = 0.91), while biexponentially derived ADC values correlated moderately well with the FS ADC (R2 = 0.51). LSDI gives reproducible ADC measurements in the lumbar discs, largely independent of b-value and signal-to-noise ratio (SNR) when used with noise correction and FS.