Reliability and variability in the interpretation of lumbar high intensity zone

Objective: The aim of this study was to evaluate the reliability of high intensity zone (HIZ) and to assess discrepancy in the interpretation, as well as investigate the effects of parameters of HIZ on interobserver variation. Methods: Four spine surgeons made independent observations on lumbar magnetic resonance imaging (MRI) from 207 consecutive patients from 3 institutions. The kappa statistic was used to characterize inter- and intraobserver reliability for visual assessments of HIZ. The corresponding MRI was provided to 2 additional spine surgeons for quantitative measurements. The parameters of HIZ, including signal intensity (SI) and area ratio (HIZ%), were used to assess the interobserver variation of HIZ. Results: The overall interobserver agreement for visual assessments was substantial (kappa=0.62 at L4-5 and 0.61 at L5-S1), and intraobserver agreement was excellent (kappa=0.84 at L4-5 and 0.86 at L5-S1). Of 93 observed HIZ, 17 instances (18.3%) were agreed upon by all visual observers. The SI with full agreement was significantly brighter than all the others (p<0.01). The HIZ% with 2 agreements was significantly smaller than those with 4 agreements (p=0.04) and 3 agreements (p=0.03). Although fewer observers with consensus were associated with smaller HIZ%, the difference was not significant (p>0.05). Conclusion: The reliability in the interpretation of HIZ was sufficient for spine surgeons with differing levels of experience. This study highlighted that signal intensity was the primary cause of variability in visual observation.