Figure 3. Local dynamic stability: descriptive statistics and comparisons between healthy controls and patients.

Forty healthy individuals (Controls) and 83 patients exhibiting mild to moderate neurological disorders (Patients) walked 2×30 sec. at preferred speed. A 3D accelerometer recorded the trunk acceleration in medio-lateral (ML), vertical (V) and antero-posterior (AP) directions. Local dynamic stability was evaluated by computing the rate of the average divergence among nearby trajectories in a reconstructed state-space that reflects the dynamics of locomotion (Lyapunov exponent method). The divergence was computed either over one step (λ_0.5) or over one stride (λ₁). The results of the two 30sec. walking bouts were averaged. The spread of the data among participants is presented with boxplots (median, quartiles and outliers (red +)). Mean and SD are shown at the bottom of the individual panels. The bold values are as follows: top: the relative difference (RD) between controls and patients, expressed as percentage; middle: effect size (ES), i.e. mean difference normalized by SD of the controls (Glass's delta); p: p-value of the t-test between controls and patients (unequal variance).