The apparent diffusion tensor (ADT) imaging method was extended to account for multiple diffusion components. A biexponential ADT imaging experiment was used to obtain separate images of rapidly and slowly diffusing water fractions in excised rat spinal cord. The fast and slow component tensors were compared and found to exhibit similar gross features, such as fractional anisotropy, in both white and gray matter. However, there were also some important differences, which are consistent with the different structures occupying intracellular and extracellular spaces. Evidence supporting the assignment of the two tensor components to extracellular and intracellular water fractions is provided by an NMR spectroscopic investigation of homogeneous samples of brain tissue. An underlying assumption for the apparent diffusion tensor (ADT) imaging method, as implemented to date, is that tissue water can be represented by a single diffusing component within each image pixel. The ADT method of Basser et al. (1,2) is an extension of the monoexponential isotropic diffusion model of Stejskal and Tanner (3), which takes into account diffusion anisotropy. NMR imaging and spectroscopic studies have shown, however, that at high diffusion weighting (b Ͼ 1500 s/mm 2 ) cells and tissues exhibit signal decay that is not monoexponential and may reflect the extracellular and intracellular water compartments (4 -12). The consequences of multiexponentiality on apparent diffusion coefficient (ADC) and ADT imaging have been recognized (8), but separate images of distinct water compartments have not been reported.In this work the ADT formalism of Basser et al.(1,2) is extended to account specifically for biexponential diffusion (13). For monoexponential diffusion, a series of diffusion-weighted images may be used to compute a singlerate ADT image by fitting to:where b is the diffusion weighting factor, S is the b-dependent signal intensity, S 0 is the (T 2 -weighted) signal intensity in the absence of diffusion-weighting gradients, b ij is the matrix of diffusion-weighting terms (14), D ij is the i,j th element of the ADT, and i,j ϭ x,y,z. For a two-compartment model under the assumption of slow exchange (i.e., minimal water exchanges compartments between excitation and signal reception), Eq.[1] may be expanded to a linear combination of observable signal components (15):where S 0f and S 0s are the (T 2 -weighted) signal intensities in the absence of diffusion weighting gradients and the subscripts f and s denote fast and slow components of diffusion, respectively. The suitability of a biexponential ADT analysis was first tested using ADT spectroscopy on homogeneous samples of nervous tissue. Spectroscopic experiments simplified data acquisition and facilitated direct comparisons of volume fractions determined by biexponential ADT and biexponential T 2 methods. Similar compartments have also been observed using multiexponential analyses of T 2 (16 -18). Next, biexponential ADT imaging was applied to a sample of fixed normal rat spinal cord. T...