Criterion Validity of Wrist-Worn IMU-Derived Smoothness Metrics During Functional Upper-Limb Tasks

Wearable inertial measurement units (IMUs) may facilitate quantitative assessment of upper-limb motor function outside laboratory settings; however, the validity of IMU-derived movement smoothness metrics compared with optical motion capture remains unclear. This study aimed to investigate the criterion validity of wrist-worn IMU-derived smoothness metrics during functional upper-limb tasks. Twenty-seven healthy adults (14 females, 13 males) performed two functional tasks, Reaching and Hand-to-mouth, with both upper limbs. Movements were simultaneously recorded using a marker-based motion capture system and two wrist-worn IMUs, both sampling at 100 Hz. Acceleration profiles were obtained from MoCap-derived wrist trajectories and IMU signals, and four smoothness metrics were calculated: Spectral Arc Length (SPARC), normalized root mean square (nRMS), Normalized Jerk (NJ), and Log Dimensionless Jerk (LDLJ). Criterion validity was assessed through Pearson correlation, estimation error (RMSE and MAE), Bland-Altman analysis, and proportional bias assessment. During the Reaching task, IMU-derived metrics showed significant correlations with MoCap values for all indices (r = 0.41-0.74), with NJ showing the strongest association. SPARC showed the smallest mean bias between systems. During the Hand-to-mouth task, significant correlations were observed only for SPARC and NJ (r = 0.39-0.47), with increased discrepancies for nRMS and LDLJ. Wrist-worn IMUs provide comparable smoothness estimates to MoCap during specific upper-limb tasks, but agreement depends on both the selected metric and movement characteristics. These findings highlight the need for task- and metric-specific validation before translating laboratory-based movement quality assessment into wearable settings.