Balance Assessment Using a Smartwatch Inertial Measurement Unit with Principal Component Analysis for Anatomical Calibration

Abstract

Balance assessment, or posturography, tracks and prevents health complications for a variety of groups with balance impairment, including the elderly population and patients with traumatic brain injury. Wearables can revolutionize state-of-the-art posturography methods, which have recently shifted focus to clinical validation of strictly positioned inertial measurement units (IMUs) as replacements for force-plate systems. Yet, modern anatomical calibration (i.e., sensor-to-segment alignment) methods have not been utilized in inertial-based posturography studies. Functional calibration methods can replace the need for strict placement of inertial measurement units, which may be tedious or confusing for certain users. In this study, balance-related metrics from a smartwatch IMU were tested against a strictly placed IMU after using a functional calibration method. The smartwatch and strictly placed IMUs were strongly correlated in clinically relevant posturography scores (r = 0.861-0.970, p < 0.001). Additionally, the smartwatch was able to detect significant variance (p < 0.001) between pose-type scores from the mediolateral (ML) acceleration data and anterior-posterior (AP) rotation data. With this calibration method, a large problem with inertial-based posturography has been addressed, and wearable, "at-home" balance-assessment technology is within possibility.

Keywords: balance; biomechanics; calibration; coordinate system discovery; inertial measurement unit; posturography; principal component analysis; smartwatch; wearables.

Figure 1

(a) Smartwatch’s inertial coordinate system, showing the directions in which the smartwatch records acceleration, rotation rate, and gravitational field. (b) Global coordinate system, which contains the anterior-posterior (AP) direction, mediolateral direction (ML), and superior-inferior (SI) direction.

Figure 2

Images of different pose difficulties. (a) Both-legs stance; expected to be the least difficult. (b) Semi-tandem stance; expected to have moderate difficulty. (c) Single-leg stance; expected to be the most difficult.

Figure 3

Manual selection of forward flexion (FF) calibration maneuver timeseries for input into PCA algorithm. The arrow bar above the data indicates the timeseries data that were identified as the forward flexion calibration maneuver. The noise on the left is from the chest-tap calibration maneuver. The sine wave to the right of the forward flexion maneuver is the lateral bending maneuver. The order of the calibration maneuvers was the same every time, so the researcher knew which maneuver was which.

Figure 4

84 RMS AP rotation rate scores between smartwatch versus SPIMU. A positive trendline was produced with every calibration method.

Figure 5

84 RMS ML rotation rate scores between smartwatch versus SPIMU. A positive trendline was produced.

Figure 6

84 RMS 2D rotation rate scores between smartwatch versus SPIMU. A positive trendline was produced.

Figure 7

84 Smartwatch RMS AP acceleration scores versus force-plate RMS AP COP velocity scores. A positive trendline was produced with every calibration method.

Figure 8

84 Smartwatch RMS ML acceleration scores versus force-plate RMS ML COP velocity scores. A positive trendline was produced with every calibration method.

Figure 9

84 RMS 2D acceleration scores for smartwatch versus force-plate RMS 2D COP velocity scores. A positive trendline was produced.

Figure 10

RMS acceleration score bar graphs. Bar graphs show sample mean; error bars indicate one standard deviation. Significance level was corrected to α_e = 0.0026. RMANOVA test results were significant for ML (p < 0.001), 2D (p = 0.002), and 3D (p < 0.001) scores. * = significant variation in the mean scores by pose type. ** = significant equal variation in the mean scores by pose type.

Figure 11

RMS rotational velocity score bar graphs. Bar graphs show sample mean; error bars indicate one standard deviation. Significance level was corrected α_e = 0.0026. RMANOVA test results were significant for AP scores only (p < 0.001). ** = significant equal variation in the mean scores by pose type.

Figure 12

Force plate’s RMS ML COP velocity bar graphs. Bar graphs show sample mean; error bars indicate one standard deviation. Significance level was corrected to α_e = 0.0026. RMANOVA results showed significant variation in the mean scores by pose type, but sphericity was violated even in best case (Factor DOF = 1.10, F = 83.23, p < 0.001). * = significant variation in the mean scores by pose type.

Figure 13

1 Sample of before and after calibration of smartwatch rotational movement. Rotation-based maneuvers are shown (i.e., first forward flexion and then lateral bending). The one-dimensional maneuvers in the global coordinate system after calibration can be clearly seen (i.e., first ML rotation then AP rotation), whereas in the inertial system, the maneuvers are an unobservable three-dimensional movement. The AP rotation of the lateral bending maneuver was more convoluted than the ML rotation of the forward flexion maneuver; apparent secondary and tertiary rotation occurred during this potentially more difficult calibration maneuver.