Mapping Orientation Distribution Function with Spherical Encoding
暂无分享,去创建一个
C. Lin, W. I. Tseng, L. Kuo, V. J. Wedeen, J. Chen Interdisciplinary MRI/MRS Lab at Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan, Center for Optoelectronic Biomedicine, National Taiwan University College of Medicine, Taipei, Taiwan, MGH Martinos Center for Biomedical Imaging, Harvard Medical School, Charlestown, MA, United States Synopsis DSI reflects the distribution of fiber orientations by mapping of 3D probability density function and its spherical projection of proton diffusion by q-space diffusion MRI. However, DSI scanning currently requires 515 diffusion-encoding directions, it is very time consuming. Here we developed a spherical encoding method to obtain the spherical projection directly. We found that the salient orientation of fibers can be extracted from summation of the product of diffusion echo signal and an angle function. This method saves at least 50% of scanning time than that needed in DSI and has an accuracy of –1.33±6.79 in defining crossing fibers. Introduction Multi-fiber orientations within a voxel have been acquired by high angular resolution diffusion weighted imaging (HARD) [1,2] and diffusion spectrum imaging (DSI) [3]. HARD is a method to get the fiber orientation distribution under the assumption that the distribution is the superposition of multiple tensors. The method is still based on the Gaussian tensor model. DSI bypasses the Gaussian assumption and obtains the distribution of fiber orientations by directly obtaining the mapping of 3D probability density function (PDF) and its spherical projection of proton diffusion, i.e. orientation distribution function (ODF), with q space diffusion MRI. Here we developed a spherical encoding method to obtain the ODF directly without the Gaussian assumption. This method also alleviates the difficulty of DSI data acquisition, which is time consuming and requires stringent gradient performance. We used the same high angular spherical encoding as that in HARD but calculated ODF by spherical Fourier transform. We found that the salient orientation of fibers can be extracted from the summation of the product of echo signal and an angle function. Our results show that only half of the scanning time is needed in our proposed method compared with that in DSI and an accuracy of –1.33 ± 6.79 was achieved when defining fibers at 90-crossing in our phantom. Materials and Methods A phantom model was designed to simulate intersecting fibers. It comprises sheets of parallel plastic capillaries with inner and outer diameters of 50 μm and 350 μm, respectively. The capillaries were filled with water and sheets of two different orientations were stacked on each other in an interleaved fashion [4]. Images were acquired using stimulated echo diffusion sequence with 253 diffusion-encoding directions in a 5 fold-tessellated icosahedral sphere in the q space [1]. The in-plane resolution is 0.7 mm and the thickness is 4 mm. Diffusion gradient has intensity = 85 mTm, duration = 5 ms, diffusion time = 250 ms, yielding diffusion sensitivity = 3211 s mm.
[1] P. Callaghan. Principles of Nuclear Magnetic Resonance Microscopy , 1991 .
[2] Ronald N. Bracewell,et al. The Fourier Transform and Its Applications , 1966 .
[3] H. Pfeifer. Principles of Nuclear Magnetic Resonance Microscopy , 1992 .