Assessment of sex-related differences in cardiovascular control using high-order feature importance

Abstract Understanding sex-related differences in cardiovascular autonomic regulation remains a critical challenge in precision medicine, as physiological variability between males and females may impact diagnosis and treatment. Recent advances in information-theoretic approaches have enabled the characterization of complex interactions within autonomic control by quantifying the unique, redundant, and synergistic contributions of physiological features, moving beyond traditional pairwise analyses. In this study, we investigated sex-related differences in cardiovascular autonomic regulation of healthy subjects at rest, by applying a recently proposed framework for high-order feature importance to indices extracted from heart rate and blood pressure variability time series. Our results revealed stronger synergistic interactions within cardiac and vascular control loops than between them, confirming their partial functional independence. This finding highlights the importance of integrating both sources of information to capture a more comprehensive and sex-specific profile of cardiovascular dynamics. A classification model based on support vector machines was then implemented using the most synergistic features, in order to identify physiological patterns in heart rate and blood pressure variability able to distinguish between female and male subjects. Training and testing the classifier on independent datasets allowed to obtain an overall classification performance reaching approximately 69%, with higher values of Recall and F-score metrics for females than males, consistent with prior evidence suggesting more stable autonomic patterns in women. By identifying the most informative physiological features for sex classification and disentangling their unique and synergistic contributions, this study provides new insights into the structure of autonomic control across sexes. These insights into sex-specific autonomic signatures may support the development of tailored monitoring tools and sex-sensitive interventions in cardiovascular medicine.