Pitfalls of magnetic resonance imaging of alar ligament

Abstract

An observational study of variations in the appearance of the alar ligament on magnetic resonance imaging (MRI) and the normal range of lateral flexion and rotation of the atlas was performed to validate some of the premises underlying the use of MRI for the detection of injuries to the alar ligament. Fifteen healthy volunteers were included. Three sets of coronal proton-density images, and axial T2-weighted images of the craniovertebral junction, were obtained at 0.5 T with the neck in neutral position and laterally flexed (coronal proton density) or rotated (axial T2). Five of the subjects also underwent imaging at 1.5 T. The scans were independently examined twice by two radiologists. The presence of alar ligaments was recorded and a three-point scale used to grade the extent of hyperintensity exhibited by the structures: the ligament were graded as 2 and 3 if, respectively, less or more of its cross-section was hyperintense, whereas grade 1 represented a hypointense ligament. The effect of lateral flexion on image quality was assessed. Concordance analysis of the data were performed before and after dichotomising the data on grading. The atlanto-axial angle and rotation of the atlas were measured. The magnitude of movement to right was normalised to that to the left to give, respectively, the flexion index and the rotation index. The alar ligaments were most reliably seen on coronal proton-density scans, with a Maxwell's RE of 0.96 as compared with 0.46 for sagittal images. Flexion of the neck improved definition of the ligaments in only rare instances. Inter-observer disagreement was marked with respect to grading of the ligament on both coronal [composite proportion of agreement (p0)=0.44; 95% confidence intervals: 0.26, 0.64)] and sagittal scans [p0=0.40 (0.19, 0.63)]. Dichotomising the data did not appreciably improve reliability [Maxwell's RE: -0.11 (coronal scans), -0.20 (sagittal scans)]: for ligaments which demonstrated hypertensive areas (grades 2 and 3) there was complete lack of agreement for both coronal [p2=0 (0, 0.25)] and sagittal scans [p2=0 (0-0.30)]. A large response bias was found in the reports of both readers albeit in opposite directions. There was poor concordance between scans obtained at different field strengths [RE (coronal images)=0.25; RE (sagittal images)=0.14). Mean flexion index and mean rotation index were 1.00 (SD 0.03) and 1.01 (SD 0.06), respectively. The MR imaging may not be the investigation of choice for the investigation of subtle injuries to the alar ligament. Whether it can be substituted by kinematic assessment of the occipito-atlanto-axial complex with MRI warrants investigation.