Diffusion Spectrum Imaging Of Fiber White Matter Degeneration

Introduction White matter degeneration poses diagnostic problems in a variety of pathologies such as stroke, post-operative changes, trauma, ALS, MS and various psychological disorders. Diffusion tensor imaging (DTI) of normal white matter has been shown to reveal structure of the major white matter fascicles (1, 2). In white matter degeneration, however, DTI loose structural information and anisotropy (3-5). DTI is therefore frequently ambiguous in determining the underlying pathophysiology of white matter degeneration (6) .As white matter degeneration is tractspecific and voxels may contain multiple fiber populations (7), we may hypothesize that white matter degeneration may cause significant but not complete loss of intravoxel structure. A new diffusion imaging method, Diffusion Spectrum Imaging (DSI) (8), directly measures the three-dimensional spin diffusion probability distribution. Using the orientational contrast in DSI intraxoxel fiber heterogeneity can be resolved, i.e., intravoxel fiber crossing in complex architecture. This characteristic increases the possibility of detecting white matter structural changes, not only in areas of known well-defined white matter anatomy, but also in areas of complex white matter anatomy. We used diffusion spectrum imaging (DSI) to test the hypothesis that the diffusion spectrum shows residual structure in areas of Wallerian degeneration, in which DTI is featureless. Further, that fiber specific degeneration causes reduction or suppression of specific angular components of the diffusion spectrum in an anatomically meaningful way. Methods Diffusion spectrum imaging (DSI) is a version of 3D q-space diffusion imaging, which measures the microscopic three-dimensional spin diffusion function for each imaging voxel. DSI were obtained in healthy subjects and patients on a 3T Siemens Allegra scanner. The DSI sequence builds on an ECG gated twice refocused SE EPI sequence (9), which minimize the effects of pulsatile deformation and eddy-currents. Sequence parameters TE = 165 ms, TR = 3RR's and spatial resolution (3mm)3. Six to nine slices were acquired (over 3RR intervals) in 1200 heartbeats giving an approximate imaging time of 20 min. Analogous to Fourier encoding of position, the Fourier encoding of displacement within a three dimensional grid gives the diffusion spectrum. The spin displacements are encoded by means of auxiliary bipolar gradient pulses whose sensitivity is given by a vector q=γδg, where g is the diffusion-encoding gradient, δ the time duration of the gradient, and γ is the gyromagnetic ratio. In total, 437 diffusion sensitizing q-vectors were used to describe a uniform threedimensional grid within a sphere. The diffusion sensitization corresponds to b-values from 0 to 20.000 s/mm2 with isotropic spectral resolution (2qmax)-1=4.5μ and field of view (2qmin)-1=45μ. The spectra were obtained by 3D Fourier Transform of the data (Fig 1), and displayed by graphical icons, orientation density functions, representing the spectral orientations. In addition the data were fitted to a Gaussian model (diffusion tensor). As shown schematically in Fig 1, the orientational maxima of the diffusion spectra (c,d) coincide with the orientation of the fibers (a), whereas the tensor fit (e) gives a mean fiber direction in between the two true orientations (c).