Non-gaussian diffusion evaluation of the human kidney by Pade exponent model

PurposeTo evaluate the feasibility of renal diffusion quantification using the Pade exponent model (PEM) in healthy subjects. Materials and MethodsDiffusion measurements were completed in 10 healthy subjects (mean age, 32.48.9 years) on a 3T MRI scanner (Magnetom Trio, Siemens AG, Germany). A respiratory-triggered echo planar imaging sequence (15 slices with 6mm thickness; 16 b-values [0-750 s/mm(2)]; three diffusion directions; field of view: 400x375mm; Matrix 192x192; repetition time/echo time: 3000/74 ms) was acquired in the coronal direction. Parameter maps were calculated for the monoexponential, biexponential, kurtosis models, and the PEM. A regression analysis using an R-2-test and corrected Akaike information criterion (AICc) was performed to identify the best mathematical fitting to the measured diffusion-weighted imaging signal decay. ResultsThe mathematical accuracy of the PEM was significantly higher than for the other three-parameter and the monoexponential model (P<0.05), which enables more precise information about the deviation of the Gaussian behavior of the diffusion signal by the PEM. The biexponential model showed better fitting to the diffusion signal (medullar R-bi(2) 0.9890.008, AICc(bi) 113.36.6; cortical R-bi(2) 0.9920.006, AICc(bi) 113.3 +/- 5.2) than the three-parameter models (medullar R-Pade(2) 0.965 +/- 0.016, AICc(Pade) 122.6 +/- 6.4, R-K(2) 0.954 +/- 0.019, AICc(K) 128.5 +/- 6.0; cortical R-Pade(2) 0.989 +/- 0.005, AICc(Pade) 116.3 +/- 4.4, R-K(2) 0.985 +/- 0.007, AICc(K) 120.4 +/- 4.8). The monoexponential model fits least to the diffusion signal in the kidney (medullar R-mono(2) 0.898 +/- 0.039, AICc(mono) 141.4 +/- 5.6; cortical R-mono(2) 0.961 +/- 0.013, AICc(mono) 135.4 +/- 4.8). ConclusionThe PEM is a novel promising approach to quantify diffusion properties in the human kidney and might further improve functional renal MR imaging.