Improving apparent diffusion coefficient accuracy on a compact 3T MRI scanner using gradient nonlinearity correction

Gradient nonlinearity (GNL) leads to biased apparent diffusion coefficients (ADCs) in diffusion-weighted imaging. A gradient nonlinearity correction (GNLC) method has been developed for whole body systems, but is yet to be tested for the new compact 3T (C3T) scanner, which exhibits more complex GNL due to its asymmetrical design.

To assess the improvement of ADC quantification with GNLC for the C3T scanner.

Phantom measurements and retrospective analysis of patient data.

A diffusion quality control phantom with vials containing 0-30% polyvinylpyrrolidone in water was used. For in vivo data, 12 patient exams were analyzed (median age, 33).

Imaging was performed on the C3T and two commercial 3T scanners. A clinical DWI (repetition time [TR] = 10,000 msec, echo time [TE] = minimum, b = 1000 s/mm² ) sequence was used for phantom imaging and 10 patient cases and a clinical DTI (TR = 6000-10,000 msec, TE = minimum, b = 1000 s/mm² ) sequence was used for two patient cases.

The 0% vial was measured along three orthogonal axes, and at two different temperatures. The ADC for each concentration was compared between the C3T and two whole-body scanners. Cerebrospinal fluid and white matter ADCs were quantified for each patient and compared to values in literature.

Paired t-test and two-way analysis of variance (ANOVA).

For all PVP concentrations, the corrected ADC was within 2.5% of the reference ADC. On average, the ADC of cerebrospinal fluid and white matter post-GNLC were within 1% and 6%, respectively, of values reported in the literature and were significantly different from the uncorrected data (P < 0.05).

This study demonstrated that GNL effects were more severe for the C3T due to the asymmetric gradient design, but our implementation of a GNLC compensated for these effects, resulting in ADC values that are in good agreement with values from the literature.

4 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;48:1498-1507.