How do visual and smartphone camera-based shoulder ranges of motion compare?

Area of Science:

• Biomechanics
• Orthopedics
• Medical Technology

Background:

• Objective assessment of shoulder range of motion (ROM) is critical for evaluating interventions and guiding rehabilitation.
• Traditional goniometry is the standard, but visual estimation is often used due to practicality, despite lower reliability.
• Smartphone-based 2D pose estimation has emerged as a potential objective alternative for ROM assessment.

Purpose of the Study:

• To compare 2D-pose-based shoulder ROM assessment with visual estimation.
• To examine inter-observer agreement between 2D-pose estimation and visual assessment by orthopedic surgeons.

Main Methods:

• Seventeen healthy individuals underwent active ROM assessment for abduction, flexion, extension, external rotation (ER), and functional internal rotation (FIR).
• Shoulder ROM was estimated using 2D videos from three smartphones and by two orthopedic surgeons visually.
• Mixed effects models and smallest detectable difference were used to assess consistency and agreement between methods.

Main Results:

• 2D-pose and visual estimates showed excellent consistency for abduction (R²=0.99), flexion (R²=0.95), and ERII (R²=0.86).
• Good to fair consistency was found for extension (R²=0.69), ERI (R²=0.73), and FIR (R²=0.52).
• Discrepancies were noted at specific ROM levels and between observers, with visual estimates sometimes differing significantly from 2D-pose estimates.

Conclusions:

• 2D-pose-based shoulder ROM assessment is consistent with visual estimates for most movements, indicating its potential clinical utility.
• The technology may reduce observer variation, offering a more objective and reliable method for shoulder ROM evaluation.
• Further refinement is needed for specific movements like ERI and FIR to improve accuracy and consistency across all assessments.