1526 - Hemiplegic Shoulder Pain and AI: A Case Study of Bedside Biomechanics

A 58yo male, RHD, with history of osteoarthritis, subacute stroke, and left hemiparesis, presented to clinic with left hemiplegic shoulder pain (HSP).  Aching  left lateral shoulder pain started shortly after stroke onset 3-months prior.  As an active musician, he had completely stopped performing.  Pain worsened with activity and side-sleeping.  Minimal relief with NSAIDS, topicals, and therapy.  He denied prior procedure/trauma. Physical exam revealed limited active ROM of left shoulder abduction and intact finger individuation of left hand.  Left shoulder radiographs revealed glenohumeral joint space narrowing and enthesopathy at greater tuberosity.  Differential diagnoses included rotator cuff tendinopathy, arthritis, and subacromial impingement syndrome.  Prior to and following subacromial corticosteroid injection, patient was asked to perform two trials of the drinking task activity while being video recorded (< 1-minute total video).  By applying a subfield of AI known as computer vision, we extracted kinematic metrics of the drinking task from video.

Discussions: Follow-up assessment was notable for patient-reported improved pain, return to music, and significant improvement in kinematic performance of the drinking task.  HSP is a common complication after stroke associated with reduced quality of life. Several factors contribute to HSP including many mechanical factors such as glenohumeral subluxation, tendinopathies, and arthritis.  These factors are prevalent in many individuals with HSP.  Recent studies have revealed ultrasound evidence of tendinitis in the supraspinatus and biceps brachii in 2/3 of these individuals.  When treating HSP, clinicians often gauge patient response by using relatively subjective and coarse methods (e.g. visual analog scale, functional ratings, patient report).  While more objective continuous measures exist (e.g. biomechanical metrics), these measures are often regarded as impractical due to the equipment and expertise required. 

Conclusions: In a case of HSP, a bedside biomechanical assessment made possible by computer vision can be used as an objective continuous measure of treatment response.