MR imaging artifacts of the axial internal anatomy of the cervical spinal cord.

Transverse scans of the spinal cord routinely demonstrate signal variations related to the internal anatomy of the cord that do not accurately conform to histologic cross sections. This study evaluates the MR appearance of the axial anatomy of the spinal cord and provides correlation to histologic sections as a means to understand this discordance so that disease can be recognized more readily. Short TR/TE spin-echo studies, cardiac-gated multiecho spin-echo studies, and gradient-refocused-echo studies of normal excised human spinal cords, a normal volunteer, and gelatin phantoms were obtained by using the same imaging parameters at 1.5 T. Imaging artifacts were further investigated by using both a 128 x 256 and 256 x 256 matrix with a varying phase-encoded axis. Histologic sections of the excised cords, which were stained for myelin, iron, and cell bodies (Nissl), were used for correlation to the images. We found that significant Fourier truncation and partial-volume imaging artifacts modulated the MR display of the cord. On short TR/TE images a ring of high signal at the periphery of the cord was due to a truncation artifact. The appearance of the central portions of the gray and white matter was affected variably by partial-volume averaging depending on the matrix size. White-matter tracts of the cord were always lower in signal than was the gray matter on all pulse sequences. This finding was not due to iron deposition or CSF motion artifacts. We suspect that this probably was related to dense, longitudinal organization of spinal tracts and resultant anisotropy of water molecule motion similar to that seen in the pyramidal tracts, tendons, and ligaments. We recommend the use of a 128 x 256 matrix with two averages (four excitations) when obtaining axial scans of the spinal cord in living subjects. Although truncation artifacts diminish image quality, the quality is superior to that of images obtained with a 256 x 256 matrix, in which longer scanning times result in motion artifacts and reduced signal to noise.