Ability of a Set of Trunk Inertial Indexes of Gait to Identify Gait Instability and Recurrent Fallers in Parkinson's Disease

Abstract

The aims of this study were to assess the ability of 16 gait indices to identify gait instability and recurrent fallers in persons with Parkinson's disease (pwPD), regardless of age and gait speed, and to investigate their correlation with clinical and kinematic variables. The trunk acceleration patterns were acquired during the gait of 55 pwPD and 55 age-and-speed matched healthy subjects using an inertial measurement unit. We calculated the harmonic ratios (HR), percent recurrence, and percent determinism (RQAdet), coefficient of variation, normalized jerk score, and the largest Lyapunov exponent for each participant. A value of ≤1.50 for the HR in the antero-posterior direction discriminated between pwPD at Hoehn and Yahr (HY) stage 3 and healthy subjects with a 67% probability, between pwPD at HY 3 and pwPD at lower HY stages with a 73% probability, and it characterized recurrent fallers with a 77% probability. Additionally, HR in the antero-posterior direction was correlated with pelvic obliquity and rotation. RQAdet in the antero-posterior direction discriminated between pwPD and healthy subjects with 67% probability, regardless of the HY stage, and was correlated with stride duration and cadence. Therefore, HR and RQA_det in the antero-posterior direction can both be used as age- and-speed-independent markers of gait instability.

Keywords: Parkinson’s disease; accelerometry; falls; gait disorders; harmonic ratio; neurologic; postural balance; recurrence quantification analysis.

Figure 1

Graphical representation of the accelerations-derived gait indexes of a representative healthy subject: (a) amplitudes of the filtered acceleration signals in the antero-posterior (AP), medio-lateral (ML), and vertical (V) direction as a function of the time range data; (b) Harmonic Ratio values for each of the 20 considered strides; (c) 2D-reconstructed state space of the acceleration and its time-delayed copies (time delay of 12 data samples). The distance (d) of two neighboring trajectories at a one-time sample, which is needed to calculate the Lyapunov exponent, is highlighted; (d) representation of the jerks during the whole gait cycle; (e) recurrence matrix. Based on the percent of the recurrent points in the diagonal line structure parallel to the main diagonal (i.e., the blue circled points), the RQAdet was calculated.

Figure 2

(a) Graphical representation of the Harmonic Ratios in the antero-posterior, medio-lateral, and vertical directions of a representative age-and-speed-matched healthy subject (blue) and a subject with PD at Hoehn and Yahr stage = 3 (red); (b) recurrence matrices in the antero-posterior direction of the same representative subjects.

Figure 3

ROC curves for the HRs in identifying pwPD vs. HSmatched, pwPD at HY = 3 from milder HY and recurrent fallers. The red line represents the HR_AP, the blue line the HR_ML, and the green line the HR_V.

Figure 4

ROC curves of the CV (a) and RQAdetAP (b) in discriminating pwPD from HS_matched.