Analysis of b-value calculations in diffusion weighted and diffusion tensor imaging

Diffusion weighted imaging has opened new diagnostic possibilities by using microscopic diffusion of water molecules as a means of image contrast. The directional dependence of diffusion has led to the development of diffusion tensor imaging, which allows us to characterize microscopic tissue geometry. The link between the measured NMR signal and the self-diffusion tensor is established by the so-called b matrices that depend on the gradient's direction, strength, and timing. However, in the calculation of b-matrix elements, the influence of imaging gradients on each element of the b matrix is often neglected. This may cause errors, which in turn leads to an incorrect extraction of diffusion coefficients. In cases where the imaging gradients are high (high spatial resolution), these errors may be substantial. Using a generic pulsed gradient spin-echo (PGSE) imaging sequence, the effects of neglecting the imaging gradients on the b-matrix calculation are demonstrated. By measuring an isotropic phantom with this sequence it can be analytically as well as experimentally shown that large deviations in single b-matrix elements are generated. These deviations are obtained by applying the diffusion weighting in the readout direction of the imaging dimension in combination with relatively large imaging gradients. The systematic errors can be avoided by a full b-matrix calculation considering all the gradients of the sequence or by generating cross-term free signals using the geometric average of two diffusion weighted images with opposite polarity. The importance of calculating the exact b matrices by the proposed methods is based on the fact that more precise diffusion parameters are obtained for extracting correct property maps, such as fractional anisotropy, volume ratio, or conductivity tensor maps. (c) 2005 Wiley Periodicals, Inc.