Integrating machine learning and statistical techniques to evaluate postural stability during hypoxia
- 1BioSIS (Biosensing & Intelligent Systems) Lab, Centre for Intelligent Computing and Systems, University of Canberra , , ,
- 2Embry-Riddle Aeronautical University , , ,
Abstract
This study investigates the effects of hypoxic altitude on postural stability using statistical analysis and feature-selection-based machine learning. Thirty-one healthy participants were exposed to four simulated altitudes (sea level, 3,000 m, 4,000 m, and 5,000 m) in a randomized, blinded protocol within a normobaric chamber, with arterial oxygen saturation ( SpO2 ) confirming hypoxic exposure. Participants completed eight standing conditions designed to manipulate visual, vestibular, and somatosensory inputs. For each condition, 15 postural sway measures were extracted from center of pressure signals. Statistical analysis indicates that hypoxia-induced changes in postural stability are multidimensional and are most detectable under conditions that increase sensory integration demands, particularly foam based and tandem stance tasks. Moderate hypoxia increased sway variability and disrupted automatic postural control, whereas more severe hypoxia under complex task conditions elicited a constrained motor strategy. In the machine learning approach, four tree-based ensemble classifiers, including Extra Trees, Random Forest, Gradient Boosting Decision Trees, and XGBoost, were employed for altitude classification using the selected postural sway features. A feature-selection framework using Extra Trees feature importance and participant-grouped nested cross-validation was implemented. The optimized Extra Trees classifier achieved the highest accuracy of 70% using 27 selected features. The selected features highlighted the contribution of specific sway characteristics that were not fully captured by conventional univariate metrics. These findings show that moderate hypoxia, in sensory challenging environments, induces detectable multidimensional postural instability before the onset of conventionally recognized symptoms. Our results highlight the importance of task selection and multivariate analysis for detecting subtle impairments. These findings have potential applications in aerospace environments.
Related
- Beyond prediction accuracy by integrating GWAASB into wheat stability assessment across environments in Saudi Arabia
- Integrating Risk-based Quality Assurance and Predictive Process Control in Food Manufacturing: A Critical Review of Contamination Prevention, Statistical Monitoring, and Regulatory Compliance
- Smart Agriculture Through Internet of Things and Machine Learning: A Survey of Monitoring Techniques, Methods and Applications
- AntiOxiPred: Predicting Antioxidant Peptides using Machine Learning Techniques and BLAST
- Predicting Conscious Perception from Pupil's Aperture Size Using Machine Learning Techniques
BibTeX
@article{baradaran2026integrating,
title = {{Integrating machine learning and statistical techniques to evaluate postural stability during hypoxia}},
author = {Yasaman Baradaran and Abishek Shrestha and Jonathan French and Maryam Ghahramani},
journal = {Frontiers in bioengineering and biotechnology},
year = {2026},
month = oct,
doi = {10.3389/fbioe.2026.1886164},
url = {https://europepmc.org/article/PMC/PMC13719473},
note = {CC BY},
}
RIS
TY - JOUR
TI - Integrating machine learning and statistical techniques to evaluate postural stability during hypoxia
AU - Yasaman Baradaran
AU - Abishek Shrestha
AU - Jonathan French
AU - Maryam Ghahramani
PY - 2026
DA - 2026/10/16
JO - Frontiers in bioengineering and biotechnology
DO - 10.3389/fbioe.2026.1886164
UR - https://europepmc.org/article/PMC/PMC13719473
AB - This study investigates the effects of hypoxic altitude on postural stability using statistical analysis and feature-selection-based machine learning. Thirty-one healthy participants were exposed to four simulated altitudes (sea level, 3,000 m, 4,000 m, and 5,000 m) in a randomized, blinded protocol within a normobaric chamber, with arterial oxygen saturation ( SpO2 ) confirming hypoxic exposure. Participants completed eight standing conditions designed to manipulate visual, vestibular, and somatosensory inputs. For each condition, 15 postural sway measures were extracted from center of pressure signals. Statistical analysis indicates that hypoxia-induced changes in postural stability are multidimensional and are most detectable under conditions that increase sensory integration demands, particularly foam based and tandem stance tasks. Moderate hypoxia increased sway variability and disrupted automatic postural control, whereas more severe hypoxia under complex task conditions elicited a constrained motor strategy. In the machine learning approach, four tree-based ensemble classifiers, including Extra Trees, Random Forest, Gradient Boosting Decision Trees, and XGBoost, were employed for altitude classification using the selected postural sway features. A feature-selection framework using Extra Trees feature importance and participant-grouped nested cross-validation was implemented. The optimized Extra Trees classifier achieved the highest accuracy of 70% using 27 selected features. The selected features highlighted the contribution of specific sway characteristics that were not fully captured by conventional univariate metrics. These findings show that moderate hypoxia, in sensory challenging environments, induces detectable multidimensional postural instability before the onset of conventionally recognized symptoms. Our results highlight the importance of task selection and multivariate analysis for detecting subtle impairments. These findings have potential applications in aerospace environments.
SN - 2296-4185
ER -