Assessment of an Automated Method to Identify Zygapophyseal Joint Crepitus

Abstract

Objective Novel accelerometry methods have been found to accurately and reliably identify zygapophyseal (Z) joint crepitus during normal lumbar motion. However, human assessment of the recordings is extremely time consuming. This study assessed the reliability and validity of a newly developed automated method (AM) to identify Z joint crepitus. Hypothesis: The AM would provide equal accuracy in significantly less time than human observers in the assessment of Z joint crepitus. Methods: A lumbar spine was embedded in silicone in the prone position. Before pouring the silicone, tunnels were created from the posterior aspect of the Z joints to the surface. Once the spine phantom was completed, a specialized mechanical device was lowered into each tunnel and was used to strike the Z joints with a force of approximately 7 N to simulate joint crepitus. Ten accelerometers applied in a previously developed pattern were used to localize the specific origin of the joint vibrations. Each joint was struck on three different passes (n=30) while recordings from the accelerometers were made. The order of joint strikes was randomized and two human observers (HO1, HO2), as well as the AM, analyzed the oscilloscope recordings of the accelerometers to identify the joint from which the vibrations originated. This was done on two separate occasions (runs), separated by a minimum of 5 days and the order of joint strikes was re-randomized prior to the second run. Observers were blinded to the joint strikes, results of one another, and to their own results from the previous run. Intra- and inter-observer reliability and validity (actual joint struck vs. observer/automation determination from recordings) were calculated (Kw), and comparison of average time to complete analysis in seconds (TC) was also calculated (ANOVA). Results: HO1: validity Run 1 = 0.996 (95%CI: 0.990-1.0); validity Run 2 = 1.000 (1.0-1.0); intra-observer reliability (Run 1 vs. Run 2) = 0.990 (0.990-1.0); HO2: validity Run 1 = 0.992 (0.977-1.0); validity Run 2 = 0.992 (0.977-1.0); intra-observer reliability = 1.000 (1.0-1.0); AM: validity Run 1 = 0.990 (0.975-1.0); validity Run 2 = 0.990 (0.975-1.0); intra-observer reliability = 1.000 (1.0-1.0); Interobserver reliability (HO1 vs. HO2 = 0.990 (0.979-1.0); AM vs. HO1 = 0.984 (0.965-1.0); AM vs. HO2 = 0.983 (0.968-1.0). TC Results: HO1: 3,810.00 (+296.98); HO2: 2,940 (+763.68); AM: 81.50 (+21.92); ANOVA p=0.009. Conclusions: The automated method demonstrated “almost perfect” reliability and validity and was significantly faster than human assessment of Z joint crepitus in this highly controlled environment. Further testing of the automated method on human subjects is warranted.