An AI-driven video based goniometer for knee joint range of motion (ROM) Assessment: Reliability and validity compared to traditional goniometry.

BACKGROUND

Accurate measurement of knee joint range of motion (ROM) is crucial in clinical and rehabilitation settings. Traditional goniometry, which is widely used, requires calibration and is subject to human errors and measurement inconsistencies. This study introduces an AI-driven video-based goniometer and evaluates its reliability and validity compared to traditional goniometry.

METHODS

This study used a cross-sectional design in an office-based setting to test a novel video-based and AI-supported software (DETROM). ROM data for knee flexion and extension were collected from 26 healthy individuals, aged 24 to 43, with weights ranging from 52 to 124 kg, heights ranging from 1.49 to 1.92 m, and BMIs ranging from 20.32 to 36.03. Each participant's right knee range of motion (ROM) in flexion and extension was measured three times using DETROM and once using a universal goniometer (UG) by an experienced physiotherapist. DETROM utilizes a deep-learning model for pose estimation to calculate joint angles. Intrarater-reliability was assessed using the intraclass correlation coefficient (ICC), and validity was evaluated through Bland-Altman analysis.

RESULTS

DETROM demonstrated excellent reliability for knee flexion and extension, with ICC values of 0.98 and 0.87, respectively. The Bland-Altman analysis revealed a mean difference of -3.50° for flexion and 0.78° for extension between the two methods, indicating good agreement.

CONCLUSIONS

DETROM provides reliable and valid measurements of knee ROM, comparable to traditional goniometry. This technology offers a noninvasive, user-friendly, and accessible tool for clinicians and patients, potentially enhancing rehabilitation care and facilitating remote assessments.