Augmented Reality-Based Finger Joint Range of Motion Measurement: Assessment of Reliability and Concurrent Validity

Sasha Gabrielle Létourneau; Helen Jin; Ethan Peters; Ruby Grewal; Douglas Ross; Caitlin Symonette

doi:10.1016/j.jhsa.2024.10.006

Augmented Reality-Based Finger Joint Range of Motion Measurement: Assessment of Reliability and Concurrent Validity

J Hand Surg Am. 2024 Dec 11:S0363-5023(24)00514-8. doi: 10.1016/j.jhsa.2024.10.006. Online ahead of print.

Authors

Sasha Gabrielle Létourneau¹, Helen Jin², Ethan Peters³, Ruby Grewal⁴, Douglas Ross⁵, Caitlin Symonette⁵

Affiliations

¹ Division of Plastic and Reconstructive Surgery, Department of Surgery, Western University, London, Ontario, Canada. Electronic address: sletourneau@qmed.ca.
² Schulich School of Medicine and Dentistry, Western University, London, Ontario, Canada.
³ Western University, London, Ontario, Canada.
⁴ Division of Orthopedic Surgery, Department of Surgery, Western University, London, Ontario, Canada.
⁵ Division of Plastic and Reconstructive Surgery, Department of Surgery, Western University, London, Ontario, Canada.

PMID: 39665730
DOI: 10.1016/j.jhsa.2024.10.006

Abstract

Purpose: This study aimed to determine the reliability and concurrent validity of finger joint range of motion (ROM) measurement using augmented reality (AR)-based hand tracking in a sample of healthy hands. Additionally, the study aimed to determine which camera view of the hand provided ROM measurements with the highest concurrent validity at each joint.

Methods: A web application developed for smart devices using Google's MediaPipe Hands framework converted AR-generated hand landmark coordinates from camera feed into ROM angle measurements in real time for all joints. From each of five camera views, we recorded five sets of AR-based flexion and extension measurements at the metacarpophalangeal (MCP), proximal interphalangeal (PIP) and distal interphalangeal (DIP) joints of normal index to small fingers. Test-retest reliability of the five AR-based measurements in each view was evaluated as was concurrent validity of AR-based measurements relative to manual goniometry, considered the reference standard. Given accepted inter-rater reliability of manual goniometry is 10°, we considered AR-based measurements within 10° of goniometry measurements to have "acceptable" concurrent validity.

Results: Forty-eight healthy hands (median age 31, 50% left, varying ethnicities) were measured. All joints demonstrated excellent test-retest reliability (intraclass correlation coefficient >0.75) in all views in flexion and ≥2 views in extension. AR-based flexion measurements were within 10˚ of goniometry in ulnar views of the index MCP, PIP, and DIP; the long MCP and PIP; and the ring PIP and DIP. In extension, multiple views at each joint consistently yielded AR-based measurements within 10° of goniometry.

Conclusions: AR-based measurement has high concurrent validity and reliability; however, optimal camera views vary joint to joint. Validation in pathologic hands is required.

Clinical relevance: Given its excellent reliability, AR-based measurement has potential for use in monitoring changes in finger ROM after intervention, either by clinicians in-person or by patients performing remote measurements independently.

Keywords: Augmented reality; machine learning; range of motion measurement; surgical innovation; telemedicine.