Signal-based assessment of postural stability using ECG, EMG, and center of pressure measures

Objective.Postural stability reflects the integrated function of autonomic, neuromuscular, and postural control systems and deteriorates with aging and reduced physical activity. This study compared the relative measurement sensitivity of electrocardiography (ECG)-, electromyography (EMG)-, and center of pressure (COP)-derived physiological features for signal-based assessment of postural stability across static and dynamic task conditions.Approach.A total of 110 women were classified into young (YG,n= 30), active older (AOG,n= 57), and non-active older (NOG,n= 23) groups. Resting ECG, grip-force EMG during maximal voluntary contractions, and COP signals during side-by-side stance, semi-tandem stance, and the five-times sit-to-stand (FTSS) task were collected. Time-domain HRV features were extracted for ECG, frequency-domain features were extracted for EMG, and sway-related features were extracted for COP. Group differences were evaluated using statistical analyses. Machine-learning models were employed as analytical tools to quantify the relative measurement sensitivity of modality- and task-specific feature sets, rather than as diagnostic classifiers.Main results.Physiological features derived from all three modalities demonstrated significant group-related differences (p< 0.001). COP-derived features exhibited pronounced task dependency, with substantially greater sensitivity during the dynamic FTSS task compared with static stance conditions. When measurement sensitivity was quantified using machine-learning-based discriminative capacity, COP features obtained during FTSS showed the highest sensitivity, followed by COP features during semi-tandem stance, ECG-derived HRV features, and EMG frequency-domain features. The best-performing models demonstrated high classification performance, with AUC values up to 0.983 and accuracy up to 0.917.Significance.The measurement sensitivity of signal-based postural stability assessment is strongly influenced by both physiological signal selection and task design. COP-derived features, particularly those obtained during dynamic functional tasks such as FTSS, provide the highest sensitivity to functional variation, while ECG- and EMG-derived features offer complementary information on autonomic and neuromuscular contributions. These findings support a measurement-oriented framework for selecting appropriate physiological signals and task conditions for quantitative postural stability assessment in aging populations.