Characterizing the nonlinear dynamics of the human postural sway response to visual stimuli

A clear understanding of how visual information affects postural sway is crucial for assessing normal balance control and developing diagnostic and rehabilitation methods for balance disorders. However, a quantitative model of sway responses to visual perturbations with improved accuracy is still needed. We used virtual reality to apply rotational visual perturbations (0.04-1 Hz, 2.5°-15°) to 14 healthy adults. Participants were splinted at the knee and hip to ensure the ankle strategy was used. Postural responses, including body angles and ankle torques, were recorded. Initial analysis demonstrated that right-eye dominant subjects showed more coherent body sway responses, possibly related to the higher magnitude of the optical flow in the right half-plane of the visual field. Detailed analysis was therefore focused on eight subjects with large, coherent responses. A detrending method was applied to angles and torques based on the inverse Fourier transform to remove frequencies below the smallest stimuli frequency. Our methodology yielded a model with improved accuracy between the visual input and body angle output, that is, coherence values close to 1. Frequency response analysis revealed a low-pass gain characteristic and a linear phase decrease showing a consistent delay in the system across all amplitudes. A parametric model fitted to the frequency response yielded a delayed, second-order, low-pass transfer function. The transfer function gain decreased with increasing stimulus amplitude, demonstrating a nonlinear response reflecting reduced responsiveness to larger visual amplitudes. In conclusion, this paper provides an experimental and analytical framework to accurately quantify the nonlinear dynamics of postural responses to visual stimuli.<b>NEW & NOTEWORTHY</b> This paper introduces a framework to quantify nonlinear postural responses to visual stimuli. Leg splints constrained movement to enforce ankle strategy. Right-eye dominant subjects showed more coherent responses, possibly related to stronger optical flow in the right half of the visual field. To address nonstationarities, we applied inverse Fourier-based detrending, improving model accuracy. Our methodology achieved a novel nonlinear, delayed, second-order low-pass model with high accuracy, distinguishing it from existing models in the literature.