Leveraging intrinsic non-sinusoidal patterns to infer search behavior to predict exposure to respiratory stressors

This study introduces a novel entropy-based methodology to quantitatively characterize nonlinear transient breathing dynamics under respiratory stress. Environmental and pathophysiological stressors can disrupt the respiratory system's gas exchange, leading to compromise and compensatory mechanisms. We present a data-driven approach that systematically evaluates classical respiratory features alongside novel entropic features as key indicators under respiratory stress. We demonstrate that conventional metrics like breathing rate (<i>B</i>_R), time of inspiration (<i>T</i>_I), and expiration (<i>T</i>_E) fail to capture discriminating features needed to detect early ventilatory instability and predict intervention needs. An exhaustive analysis of key respiratory fiducial points using entropic methods led to novel features for understanding respiratory mechanics and classifying respiratory states. We found that the nonlinear dynamics of the transition times between inspiratory and expiratory phases (interphases) are crucial for assessing adaptability to respiratory challenges. This metric quantifies the complexity of transition duration (acceleration and deceleration between phases) and is essential for predicting declining breathing states. Our predictive model incorporating these novel approaches showed superior discriminating ability over models using classical features, achieving a 50.76% increase in predictive power as measured by the area under the curve (AUC). These findings underscore the effectiveness of this entropy-based approach for early detection of respiratory compromise, with the best model achieving an AUC of 0.784. The results have significant implications for improving clinical monitoring of acute respiratory failure and managing chronic respiratory conditions.<b>NEW & NOTEWORTHY</b> Entropy-based metrics analyzing respiratory phase transitions (inspiration-to-expiration and expiration-to-inspiration) detect respiratory compromise under hypoxic conditions better than standard breathing rate measurements. Analysis of nonlinear dynamics during these transitions reveals key ventilatory adaptations during exposure to respiratory stressors. Measuring timing variations at phase transitions improves predictive model performance in detecting exposure to hypoxic environments by a 50.76% increase in area under the curve (AUC) vs. classical methods.