Purpose: The purpose of this study was to evaluate the feasibility of diffusion tensor imaging of the kidney at a field strength of 3T. We assessed fractional anisotropy (FA) and apparent diffusion coefficients (ADC) of various acquisition protocols and determined the reproducibility of these measurements. FA, ADC, signal-to-noise ratios (SNR), and contrast-to-noise ratios (CNR) were compared with those acquired at 1.5T.
Material and methods: Ten healthy volunteers were examined with a respiratory-triggered echo-planar imaging sequence (TR: 1800 ms, TE: 58 ms, b = 0, 300 s/mm(2)) on a 3-Tesla whole-body MR scanner. Protocol variations included diffusion measurements during free-breathing, in 6 or 12 directions, and an additional b-value of 50 s/mm(2). A breath-hold protocol was also integrated (TR: 820 ms, TE: 58 ms, b = 0, 300 s/mm(2)). Measurements with 2 b-values and 6 diffusion directions were also acquired at 1.5 T. SNR was calculated with the difference-image method. Statistical analysis was performed with Wilcoxon signed-rank tests. Intrareader correlation was assessed with weighted kappa coefficients and reproducibility with the root-mean-square-average and the Bland-Altman-method.
Results: At 3T, SNR of cortex and medulla and CNR of cortex/medulla were significantly higher than at 1.5T, leading to improved corticomedullary discrimination. There were no significant FA- and ADC differences with 2 b-values and 6 diffusion directions between measurements at 1.5T and 3T. FA of the medulla was significantly higher than that of the cortex in all measurements. Tractography visualized a typical radial diffusion direction in the medulla. Best image quality was achieved with a respiratory triggered protocol with 12 acquisition directions. Measurements with 3 b-values led to decreased ADCs. Acquisition in 12 directions resulted in decreased cortical FA. FA and ADC of breath-hold and free-breathing acquisitions were significantly higher than that of the respiratory-triggered protocol. Intrareader correlation ranged from kappa 0.60 to 0.96. Variance of the respiratory-triggered protocol was smaller than that of breath-hold and free-breathing protocols. Variance was highest for medullary FA in all protocols with reproducibility coefficients ranging from 0.36 to 0.46.
Conclusion: Diffusion tensor imaging of the kidney at 3T is feasible and yields significantly higher SNR and CNR. FA and ADCs do not significantly differ from 1.5T. Number of b-values influences ADC-values. Acquisitions in 12 directions provide lower cortical FA-values. We recommend a respiratory-triggered protocol because of improved image quality and reproducibility.