Use of image fusion in radiosurgery treatment planning can reduce systematic MR image distortion

Geometrical image accuracy in radiosurgical treatment planning is of paramount importance. Several reports have suggested that magnetic susceptibility artifacts may cause spatial distortion in MR images. We have performed a systematic evaluation of the spatial accuracy of MR images relative to X-ray CT. Sections of a 5 French angiography catheter cut to a length of 5 mm and filled with a CuSO₄ solution were used in a common CT/MR phantom; additional markers were attached to the head ring of stereotactic patients. These markers, as well as anatomical structures, were identified on multiple CT and fused and non-fused axial and coronal MR slices. Two commercial radiosurgery treatment planning systems were used to perform the image fusion; one system allows manual manipulation based on visual queues. The lateral, anterior/posterior (AP), and axial (vertical) coordinates of the center-of-mass of the fiducial markers and internal anatomical landmarks were subsequently recorded. The deviation of fused and non-fused MR relative to CT was determined in each of the three principle directions. In analyzing any random or systematic displacement, both the true mean, including negative values as well as positive, and the mean of the absolute deviation were calculated. The 3D displacement, computed as [x² +y²+z²]^1/2, was also recorded. While lateral or axial deviations were observed to be essentially random, results suggest a systematic shift in the AP dimension (posteriorally), corresponding to the frequency encoding direction on axial MR images. A comparison of fused and non-fused MR image sets suggests that image fusion can reduce this posterior shift