Validation of MRI-Derived Morphometric Estimations of Biomechanical Inputs to Improve Low Back Pain Risk Assessment

Abstract

Low back pain (LBP) is one of the most common musculoskeletal disorders facing the working world. Previous LBP studies have focused on exposure to physical risk factors in the workplace such as lifting, pushing, pulling and awkward postures. Subject personal char- acteristics can vary significantly. Biomechanically relevant structures, such as internal low back geometry, impact biomechanics and resulting forces experienced by individuals. Most biomechanical models do not consider these important differences. Often, models rely on av- erages which may over- or underestimate forces and subsequent risk. This dissertation seeks to provide better inputs to these biomechanical models by first exploring methods to accurately measure and estimate these structures. Further, this dissertation seeks to demonstrate that the incorporation of such structural information into biomechanical models yields more predictive outcomes. Several studies have observed that personal characteristics such as age and gender are predictive of LBP. However, studies that focus on the impact of internal muscle and bone configuration, particularly those that utilize MRI-derived characteristics of the lumbar struc- ture, are very rare. The aims of this research are to: 1) investigate relationships between gross anthropometric characteristics and internal low back geometry; 2) comprehensively evaluate the repeatability of MRI-based measures used to produce regression relationships for low back structures; 3) investigate novel approaches to quantifying lumbar endplate degeneration; and 4) improve the predictive ability of ergonomic models by incorporating subject specific char- acteristics. This study uses Magnetic Resonance Imaging (MRI) scans to precisely measure low back geometry. This research consists of three different studies. The first study was conducted to assess reliability of MRI scans. Thirty-six (36) subjects (20 male, mean age = 24 years ± 3.1; 16 female, mean age = 25 ± 4.7) were scanned using a 3T scanner using a standardized T2 weighted protocol. Two operators who were blinded to subject identity and scan order per- formed the scanning procedures. The sagittal view and the axial view of the lumbar spine were obtained from the subjects. Subject demographics (such as age, gender, height and weight) were also recorded. Using OsiriX (v8.0.1, 2016, Antoine Rosset, Bernex, Switzerland), software, each exam- iner measured the anterior and posterior height of the vertebrae, superior and inferior length of vertebrae, concavity level of the intervertebral disc, anterior and posterior height of the inter- vertebral disc, vertebral body width and height, Pfirrmann Intervertebral Disc Grading (PIDG), vertebral angle, disc length and the sizes of the psoas and erector spinae muscles). Inter- and intra-rater reliabilities were investigated for all of these measures. In addition, reliability for the entire process was evaluated using a worst-case scenario comparing two distinct scans of the same subject with different researchers performing each MRI scan and different researchers performing the measurement of those scans using Osirix. Subsequent analyses were conducted to evaluate intra-rater reliabilities for researchers evaluating distinct scans including their own. For the second study, T2-weighted MRI scans were obtained from fifty (50) subjects (25 females, mean age = 29 years ± 5.8; 16 male, mean age = 32 ± 4.7) who had no current low back pain or self-reported low back injury. The MRI scans contained the sagittal profile of the lumbar endplates (L2-S1). Each examiner measured the height and concavity level of each lumbar disc. These measures were used to calculate a novel metric: the Concavity Index (CI; concavity level divided by vertebral body height). CIs were compared to Pfirrmann IVD grad- ing scores to evaluate their agreement and compare their respective inter-observer reliabilities. A linear relationship between average CI and corresponding Pfirrmann classification was ob- served. While overall agreement among Pfirrmaann raters was high, 10% of ratings disagreed by two categories. There was never disagreement by more than two categories. CIs had an av- erage coefficient of variation of 0.95% across all participants and lumbar regions. This presents an alternative method for quantifying intervertebral disc degeneration that appears to have ad- vantages over the traditional Pfirrmann grading scale. Most notably, the objective, quantitative, and repeatable nature of the CI. The third study details the feasibility of incorporating personal characteristics into exist- ing ergonomic tools. The L5/S1 Intervertebral Disc (IVD) cross sectional area (estimated using regression relationship), age, gender, height and weight were explored as possible risk factors. These factors were applied as multipliers to the Revised NIOSH lifting equation (RNLE) to determine if risk estimation could be improved. These multipliers were validated using a U.S. automotive manufacturer database with known health outcomes. The odds ratios and performance of the tool using only the traditional NIOSH lifting equation multipliers were compared to the new additional multipliers. This study showed that including these personal characteristics into the RNLE improved the odds ratios significantly.