Monitoring Arm Recovery After Stroke (MARS)

Upper limb recovery after stroke remains poor and 20% of stroke survivors do not recover arm movement. To improve outcome and advance insights to recovery mechanisms, an international collaboration has proposed a standardised set of outcome measures, including movement kinematics. Kinematics are moresensitive to change than clinical measures and can differentiate whether recovery is achieved by compensating to impairment or true recovery. However, kinematic assessments are not performed in clinical practice as 3-D motion capture requires expensive equipment and expertise for set-up and analysis.

The investigators therefore aim to develop a low-cost tool to measure kinematics. Open-source Artificial Intelligence models can detect positions and orientations on video and are called pose estimation models. The objectives will be to deploy and test these models in stroke survivors. The investigators will invite 12 stroke survivors with mild to moderate upper limb impairment and compare the accuracy of the models to gold-standard kinematic analysis when performing a variety of upper limb tasks. The investigators will optimise the models in case of any discrepancies. The investigators will develop a front-end smartphone app to instruct, record and provide feedback of arm movement performance to clinicians and stroke survivors. The investigators will develop the software back-end performing analysis of recorded movements and integrating these findings into electronic healthcare records for longitudinal performance tracking.

This accessible technology will provide clinicians kinematic analyses. Kinematics can guide treatment modifications and progression to improve upper limb movement.

Stroke survivors

Inclusion criteria

• History of stroke
• Arm impairment evidenced by Fugl-Meyer Upper Limb Assessment between 9-60/66.

Exclusion criteria

• Severe cognitive impairment preventing ability to consent to treatment and understand and follow research protocol
• Severe language deficit preventing ability to consent to treatment and understand and follow research protocol
• Shoulder pain >3/10 on visual analog scale
• Unable to maintain independent sitting balance without a high back support.
• Wheelchair users that are unable to transfer with assistance of 1 to lab chair or whose wheelchair backrest cannot colapse.