| dc.description.abstract | Rotator cuff tears are a common source of shoulder pain that requires conservative
management or surgical intervention to heal and regain proper function. During both
interventions, prescribed exercise programs are given to patients as they increase range of motion
(ROM) and improve patient outcome scores. However, when tasked with completing exercises in
the home, patient adherence usually decreases and is subjectively monitored by the patient
themselves. Wearable sensor devices, such as smartwatches, demonstrate feasibility to
objectively track shoulder exercise adherence using machine learning, but these algorithms
require a broad range of training data in order to accurately classify exercise type. Further, to
monitor shoulder exercise rehabilitation, sensor training data should include compensatory
exercise performance associated with symptomatic individuals. However, capturing this
movement data from a symptomatic population presents a number of challenges. To address this
problem, the objective of this study was to determine if asymptomatic individuals can simulate
compensatory movement cues associated with subacromial impingement during various
rehabilitative shoulder exercises. Seventeen participants (10 asymptomatic and 7 symptomatic for subacromial
impingement) performed twenty repetitions of six evidence-based shoulder exercises following
standard and compensatory movement cues based on their group classification. Kinematics of
the torso and upper limbs were collected to identify changes in maximum angle and ROM for
torso, thoracohumeral and elbow joint angles. Time-series joint angle data were compared for the
standard and compensatory conditions performed by the asymptomatic group using statistical
parametric mapping (SPM). Symptomatic and asymptomatic (compensatory) were compared
using maximum angle and ROM measures. Asymptomatic participants were successful in simulating compensatory movement cues based on changes in their time-series data. Differences occurred in the middle portion of the
thoracohumeral elevation time-series profile during the flexion (p < 0.05), scaption (p < 0.05),
and abduction (p < 0.05) exercises. Further, these simulated compensatory movements were
similar to the movement patterns of some symptomatic participants. Overall, these results
suggest that asymptomatic individuals can execute both standard and compensatory movement
cues. The variability of the data collected represents a spectrum between worst-case
compensatory and best-case proper movement for the six shoulder exercises performed. Further
research is needed to better understand the range of symptomatic exercise performance in order
to refine the movement cue instructions for asymptomatic individual performance. Data and
findings from this work provide crucial groundwork towards the development of improved
machine learning algorithms for sensor-based tracking of rehabilitative shoulder exercise
program adherence and progression. | en |
