Project Summary Identifying the optimal treatment for chronic low back pain (CLBP), the most prevalent of painful musculoskeletal disorders, on a patient-specific basis is an important and unresolved challenge. Tailoring interventions according to patient movement characteristics may improve clinical outcomes. Multi-modal studies are underway to characterize CLBP patients and to provide insight into the phenotypes associated with the experience of CLBP in relation to direct targeted and improved treatments. Comprehensive assessment of lumbar spine movement CLBP patients may be an important facet of such treatments such that patient-specific spine biomechanics may be included in predictive models to improve the ability to characterize them. To that end, the purpose of this administrative supplement is to explore additional clinical tools for characterizing lumbopelvic kinematics during functional tasks and daily activities. Specifically, this work aims to develop computer vision methods with which to characterize motions of the lumbar spine, thereby providing an accurate and unobtrusive clinical tool that can be used to supplement or replace currently considered tools such as wearable sensors, handheld sensors, and complex marker-based motion capture systems. This project will use video, marker-based 3D video motion capture data, and wearable inertial measurement data to create and validate single-camera computer-vision algorithms that can be used to compute known clinical metrics. During clinical assessments of patients in the Pitt LB3P Biomechanics Core study, participants are asked to perform functional tasks (e.g., repeated flexion/extension, axial rotation, lateral bending, lifting, chair rises) while wearing inertial measurement units (IMUs) attached to the upper back at T1, low back at L1 and L5, and thigh. The functional performance exams are also recorded by video. The standard metrics determined from these trials will include maximum and minimum lumbar spine and hip ROM, angular velocity at mid-excursion, maximum rotation acceleration, and phase angles for lumbar and hip joint rotation. The goal of this supplemental proposal is to use collected data to develop and train computer vision algorithms (so-called markerless motion capture) to quantify the metrics of interest. Use of video is much simpler for clinicians as it avoids the setup process required by wearable sensors. Recent developments in markerless motion capture have enabled simple camera systems to provide quantitative information about human motions but few studies have assessed computer vision for use in the clinical setting. The research plan in this proposal involves the use of existing data to train and validate computer vision algorithms, and then to further investigate their use in the clinical setting with single camera video to determine the extent to which video-based markerless motion capture may be clinically useful in the assessment of CLBP patients.

Project Narrative Chronic low back pain (CLBP) is a common musculoskeletal disorder with major public health and socioeconomic impacts. A clinical measure that is used to characterize and ultimately prescribe treatment is the extent to which a patient has difficulty executing physical activities, but making reliable quantified measurements of patient motions can be challenging with conventional tools and even with modern motion capture systems. This research will investigate the degree to which a simple camera may be used in conjunction with machine-learning based computer vision algorithms to provide patient motion during physical activities that can be used in the assessment and treatment of CLBP.