Superiority of CT Myelography over MRI in Delineating the Spinal Cord during Spine Stereotactic Radiosurgery

Abstract

Spine stereotactic radiosurgery (sSRS) has an important role in the management of primary and metastatic spine tumors. A critical component of sSRS is the accurate delineation of the spinal cord to prevent severe and potentially irreversible myelopathy. CT myelography (myelo) and magnetic resonance imaging (MRI) registration are commonly used during sSRS planning to delineate the spinal cord. The aim of this study was to rigorously compare these two modalities and the potential impact they have on planning outcomes. Between 3/2016 and 10/2016, 24 cases of sSRS were simulated using CT myelo with a companion diagnostic MRI. Two independent and blinded physicians contoured the MRI and myelo cords. Registration was then performed to the vertebral column and spinal canal. Ninety-six treatment plans were generated to deliver either 16 Gy or 18 Gy in a single fraction to the tumor planning target volume (PTV), using either MRI or CT myelo defined cords as an avoidance structure. CT myelo cord D.01cc, cord D10%, and PTV D99% were then calculated. MRI and CT myelo spinal cord volumes were compared using two-tailed paired t-test, and Dice similarity coefficients were calculated to assess their spatial overlap (where 0 equals no overlap and 1 equals perfect overlap). Thirteen (54%) cases were post-op and 11 (46%) cases were non-operated. The mean Dice similarity coefficient was 0.68 +/- 0.11. There were no significant differences in cord volumes (mean 4.7 +/- 2.5 cc for MRI vs. 4.5 +/- 2.4 cc for CT myelo, p=0.25), suggesting that the majority of discordance was spatial rather than volumetric. For post-op cases, there was a trend towards greater cord volume on MRI compared to CT myelo (5.1 +/- 1.5 cc vs. 4.7 +/- 1.1 cc, respectively, p = 0.088). When the MRI defined cords were used for treatment planning, PTV coverage (PTV D99%) was significantly less than with myelo-based treatment planning (for 16 Gy sSRS, 13.5 vs. 14.1 cc, p = 0.008; for 18 Gy sSRS, 14.6 vs. 15.2 cc, p = 0.012). Cord doses were significantly greater with MRI planning as well. D0.1 cc for 16 Gy plans for MRI vs. CT myelo was 12.6 +/- 1.9 Gy vs. 10.4 +/- 1.5 Gy (p<0.001), and for 18 Gy plans was 13.4 +/- 2.3 Gy vs. 11.0 +/- 1.5 Gy (p<0.001). 26 of the 48 MRI plans (54%) exceeded cord dose constraints, with one patient having a D0.1 cc of 18.2 Gy (EQD2 of ∼90 Gy). There are significant spatial differences in MRI and CT myelo defined spinal cords. MRI-based planning consistently resulted in worse PTV coverage and higher doses to the spinal cord. Spinal cord doses were unacceptably high in > 50% of MRI planned sSRS cases. Our data supports spinal CT myelography as the gold-standard for spinal cord delineation, and caution should be used when relying on MRI registration to define the cord for single fraction sSRS.