Mapping the orientation of intravoxel crossing fibers based on the phase information of diffusion circular spectrum.

A new method is presented to map the orientation of intravoxel crossing fibers by using the phase of the diffusion circular spectrum harmonics. In a previous study [Zhan, W., Gu, H., Xu, S., Silbersweig, D.A., Stern, E., Yang, Y., 2003. Circular spectrum mapping for intravoxel fiber structures based on high angular resolution apparent diffusion coefficients. Magn. Reson. Med. 49, 1077-1088], we demonstrated that the magnitude of the 4th-order harmonic of the diffusion circular spectrum can be used to identify the existence of fiber crossings. However, the orientation of the intravoxel crossing fibers remained unknown. This study extends the diffusion circular spectrum mapping method so that it is able to identify the orientation of the intravoxel crossing fibers by utilizing the phase information of the circular spectrum. In general, the phase of the circular harmonic determines the rotation of the apparent diffusion coefficient (ADC) profile on the sampling circle that is spanned by the major and medium eigenvector of the diffusion tensor and thus can be used to determine the orientation of the crossing fibers. Simulation results show that the regular tensor-based major eigenvector maps have obvious artifacts in the fiber-crossing area, whereas the estimated crossing fibers by the proposed method are much more consistent with the orientation of the actual intravoxel fibers. Diffusion MRI experiments were performed on five healthy human brains using a 3T scanner. The brain regions with fiber crossings were selected by thresholding the magnitudes of the 4th-order circular spectrum map. Intravoxel crossing fibers were estimated by the phase of the 4th-order harmonic for each voxel within these areas. The estimated intravoxel crossing fibers demonstrated a clear consistency with the orientations of fiber tracks in the surrounding tissues, reducing the fiber orientation discontinuity of the regular major eigenvector map.