Determining the lumbar vertebral segments on magnetic resonance imaging

Abstract

Study design: A study to test the ability of an additional cervicothoracic localizer scan to decrease interobserver discrepancy in the identification of vertebral segments in magnetic resonance imaging of the lumbar spine.

Objectives: To investigate whether lumbar vertebral segments can be identified correctly from lumbosacral magnetic resonance localizer scans, the degree of interobserver discrepancy, and the value of an additional cervicothoracic localizer scan.

Summary of background data: In magnetic resonance imaging of the lumbar spine, it may be difficult to identify transitional lumbosacral vertebral segments.

Methods: The sagittal and coronal lumbosacral localizer scans of 141 consecutive patients referred for magnetic resonance imaging of the lumbosacral spine were reviewed independently by two radiologists with the aim of locating the L5 vertebra. An additional sagittal cervicothoracic localizer scan also was performed in each case. The final study group consisted of 129 patients. The L5 vertebra was identified by counting caudally from C2 using the sagittal cervicothoracic and lumbosacral localizer scans. In the 54 most recently studied patients, cod liver oil capsule surface markers were placed near the thoracolumbar junction to quantify any marker shift between the two sagittal localizer scans.

Results: The lumbar segments could be identified consistently by counting caudally using cervicothoracic and lumbosacral localizer scans. Using sagittal lumbosacral localizer scans alone, the lumbar vertebral segments could be identified correctly in only 80.2% of patients. Coronal lumbosacral localizer scans produced similar results (82.2%). The accuracy fell to 77.9% when using a combination of both sagittal and coronal lumbosacral localizer scans. There was a 11.6% interobserver discordance in assessment of these levels. Lumbosacral transitional vertebrae were identified in 17 patients (13.2%), including 8 sacralized L5 and 9 lumbarized S1 vertebrae. Apparent surface-marker shift between cervicothoracic and lumbosacral localizer scans was insignificant, averaging only 1.9 mm (range, 0.0-5.6 mm).

Conclusions: The addition of a cervicothoracic localizer scan in magnetic resonance imaging of the lumbosacral spine is highly recommended.