An Adaptive State- and Morphology-Based Four-Event Segmentation Algorithm for Sit-to-Stand-to-Sit Cycles from Wii Balance Board Load Signals: An Exploratory Comparison with Noraxon Report-Consistent IMU Timing

Abstract

Sit-to-stand (STS) tests are often summarized by completion time or repetition count; however, phase-specific analysis requires reproducible operational segmentation. We developed an adaptive state- and morphology-based algorithm that identified ten internal anchors in Wii Balance Board (WBB) vertical-load waveforms and derived an ordered P1–P4 chain representing cycle start, end of sit-to-stand, start of stand-to-sit, and cycle end, respectively. The algorithm was developed using 108 T1–T3 trials and then applied unchanged to a temporally separate batch of 24 T0 trials. Fourteen T0 participants overlapped with development and 10 were included in a descriptive subset of participants not represented in the development dataset. Complete chains were identified in 24/24 T0 trials and 10/10 descriptive-subset trials (100%; exact 95% CI, 69.2%–100%), versus 7/24 for the prespecified internal non-adaptive baseline. Complete-chain identification remained 100% across all three datasets under one-at-a-time perturbations of ±10% and ±20% of 52 frozen numerical parameters, although selected sit-down rules shifted local boundaries. Removing the separate seated-baseline/fallback component reduced complete-chain identification to 101/108, 12/24, and 8/10. Across 131 paired trials, WBB total-cycle duration differed from Noraxon software (MR 4.2.34 Software; Noraxon, Scottsdale, AZ, USA)-reported duration by −0.188 s (95% CI, −0.313 to −0.060). These differences reflect distinct operational timing definitions rather than error against an independent reference standard. Independent reference-event and external clinical studies remain necessary.