Accuracy of q-space derived parameters in MRI - A phantom study of system-induced limitations

Introduction The aim of this study was to compare the accuracy of parameters derived from q-space measurements between a 3.0 T clinical MRI scanner and a 4.7 T NMR spectrometer. In order to do this, measurements were performed on n-decane, at the two systems using similar pulse sequence parameters. Materials and Methods The n-decane is a linear hydrocarbon (C10H22) that gives a diffusion coefficient in the same magnitude as in vivo cerebral matter [1]. MRI measurements were performed on a 3.0 T Siemens Allegra system. A double refocused SE EPI pulse sequence was used. Diffusion encoding was performed in one direction, given by (x,y,z)=(1,1,1). The temperature was 21C. NMR measurements were performed on a 4.7 T Bruker DMX 200 MHz spectrometer equipped with a DIFF-25 gradient probe driven by a BAFPA-40 unit, at 20C using a PGSTE sequence. The diffusion sensitivity was calculated according to the Stejskal-Tanner equation, b=γδG(∆-δ/3) where Td = (∆−δ/3) is the diffusion time, δ is the pulse duration, ∆ is the time between the two pulses and G is the gradient amplitude. The q-value is defined as, q=γδG/2π [m]. MRI measurements were carried out using 45 different b values, whereas the NMR experiment used 64 different values. The signal decay curves were analyzed after noise correction without any zero filling. Diffusion coefficients (D) and kurtosis [2] were determined from the signal attenuation curve and the full width at half maximum (FWHM) value was determined from the diffusion propagator. Experimental parameters for both systems are summarized in table 1. Results In figure 1, signal decay curves from the two systems are shown. The numerical results are summarized in Table 1. The MRI system gives higher D values than the NMR spectrometer for the short diffusion times, but approaches the NMR measurements for increasing diffusion times. The differences can partially be explained by neglected cross terms from the imaging gradients, which for the highest b-values were determined to be within 5 %. The resolution in the q-space measurements is given as 1/qmax, and this will also become the limiting factor for determination of the FWHM value. The limited resolution explains why the measured FWHM values are not in agreement with the expected FWHM, as calculated from the root mean square displacement. Instead the measured FWHM values follow the resolution limits. A similar effect is seen in the kurtosis measurement, where kurtosis values from MRI approached the expected value (k=0) when q-space resolution was increased. This effect was also verified in a simulation (fig. 2) showing the numerical kurtosis values as a function of used maximum q value (resolution limit). However, adding noise to the simulation in order to reproduce a typical MRI measurement, where a signal to noise ratio (SNR) of 30 in the b=0 measurement is representative, gives an unreasonable high kurtosis if the signal is sampled to far out in the q-space.