A PREOPERATIVE COMPARISON OF SENSOR-BASED FREE-LIVING AND IN-CLINIC VIDEO-BASED GAIT MEASURES IN PATIENTS AWAITING KNEE ARTHROPLASTY SURGERY

Gait metrics have been used to predict knee osteoarthritis (OA) progression and classify OA severity, but require the use of sophisticated laboratory-based equipment1,2. Wearable sensors offer a solution to monitor free-living gait over extended periods of time, yet the relationship between sensor-based and video-based gait measures is not well understood. Therefore, the purpose of this study is to examine the association of daily walking metrics captured by an accelerometer and knee kinematic gait outcomes measured using video-based motion capture in a population of people awaiting knee arthroplasty surgery. Patients with end-stage knee OA were recruited from participating surgeons’ knee arthroplasty waitlists. Gait outcomes were measured using an in-clinic video-based, markerless motion-capture system (Sony, Theia Markerless). Data from ten gait cycles for each participant were processed to calculate knee flexion and adduction angles, knee angular velocities, and gait speed (Visual3D, C-Motion). Free-living gait was measured using a single accelerometer, secured on the shank of the affected limb of each participant (Axivity). Acceleration data reflecting a continuous one-minute walking bout occurring in the middle of the free-living period were extracted for each participant for analysis. Raw acceleration peaks were used to identify foot contact to segment participant steps3. Processed acceleration data were used to align the axes of the sensor to the anatomical axes of the limb segment, and to compute linear shank velocities in the frontal plane and the thrust accelerations (associated with knee adduction moments) during early stance4,5,6. The mean video-based outcomes during stance phase for all participants were compared to the mean sensor-obtained measurements of peak stance acceleration and velocity, range in acceleration and velocity from initial contact to the peak during stance, and thrust acceleration using Pearson's correlation coefficients (r). Video-based and free-living sensor gait data were collected and compared from 21 knee arthroplasty patients awaiting surgery (13M/8F) with an average age and BMI of 69 years (±6) and 33 kg/m^2 (±8) (Table 1). A higher peak knee flexion angle during stance phase (video) was moderately correlated with a higher frontal plane acceleration range from initial contact to peak stance (sensor), a higher frontal plane velocity range from initial contact to peak stance (sensor), and a higher thrust acceleration (sensor). A higher gait speed (video) was moderately correlated with a higher frontal plane acceleration range from initial contact to peak stance (sensor). While the sensor and in-clinic measures were not strongly correlated, the sensor data presented an opportunity to define different, free-living gait outcomes, providing complementary insight into gait mechanics. A better understanding of patient functional variability may be gained, potentially relevant to arthroplasty patient care and decision-making. Gait varies between clinical and real-world settings, meaning the development of clinically relevant metrics to quantify free-living gait mechanics using simple technologies may generate more widespread translational uptake of gait considerations into clinical decision-making. For any figures or tables, please contact the authors directly.