76 - Assessment of an Automated Method to Identify Zygapophyseal Joint Crepitus in Human Subjects

Poster Board Number: 76
There are separate poster presentation times for odd and even posters.
Odd poster #s – first hour
Even poster #s – second hour

Abstract Body : Introduction and Objective: Zygapophyseal (Z) joint crepitus during normal lumbar motion has been associated with decreased joint lubrication and potential hypomobility. Consequently, crepitus may be a biomarker of Z joint dysfunction. Although novel accelerometry methods have been found to accurately and reliably identify Z joint crepitus, human assessment of the recordings is extremely time consuming. We previously validated an automated method (AM) of assessing Z joint crepitus in a spine phantom. This current study assessed the reliability of the AM to identify Z joint crepitus in humans before and after spinal manipulative therapy (SMT). Hypothesis: The AM would provide substantial reliability in significantly less time than human observers in the assessment of Z joint crepitus. Material and Methods: This project was approved by the institutional IRB. Using previously reported reliable and valid methods to identify Z joint crepitus, 11 accelerometers were applied to the lumbar region of 12 human subjects, 6 healthy and 6 with low back pain (age 54.4 +7.1). Recordings were made from the accelerometers during flexion and extension, both before and after SMT. The 48 oscilloscope recordings of the accelerometers were analyzed to identify the joint from which crepitus vibrations originated. Analysis was conducted by two human observers (HO1, HO2) as well as the AM. Based on our previous studies and post-study analysis, the crepitus definition used by observers and the AM in this study was designed to be conservative and did not identify subtle crepitus events. The observers and operator of the AM were blinded to any subject information, whether the recordings were taken before or after SMT, and each other’s work. Inter-observer reliability (Kw) and comparison of total time to complete analysis in seconds were calculated. Results: Eight crepitus events were identified from 4 individuals (2 low back pain, 2 healthy). Five events were pre-SMT. All events were identified during flexion. Inter-observer reliability following machine learning paradigm adjusting was (Kw=0.788; SE=0.056, 95% CI=0.0.682-0.895). Time to completion results (in seconds): HO1: 14,883; HO2: 11,706; AM: 246. Conclusions: The automated method demonstrated substantial reliability and was 20-54 times faster than human assessment of Z joint crepitus in the assessment of human crepitus data. Significance and Implication: These results support the use of the developed AM in the assessment of Z joint crepitus in future studies. Further testing of the AM on larger numbers of human subjects using machine learning to improve reliability and identification of subtle crepitus is warranted.