Aspiration is prevalent in the elderly population, and can lead to life-threatening conditions such as suffocation and aspiration pneumonia. However, gold-standard diagnostic approaches for aspiration-videofluoroscopic swallowing studies (VFSS) and fiberoptic endoscopic evaluation of swallowing (FEES)-involve radiation, invasive procedures, or operational complexity, and cannot provide continuous aspiration monitoring. Here, we develop a skin-interfaced wireless biomechanical system with tailored machine learning algorithms to detect aspiration in a continuous, user-friendly, non-invasive, and automated manner. The system detects post-swallow choking coughs as an indicator of aspiration, and evaluates a nursing-home cohort including aspiration elderly (n = 14) and non-aspiration elderly (n = 64). Leave-one-out cross-validation yields an average accuracy of 92% across four activity types, with a sensitivity of 91% in aspiration recognition. Additional testing on unseen aspiration patients supports patient-level generalization. Reliability analysis and confidence calibration enhance the interpretability of the machine learning classification model, and underpin clinical deployment. The approach provides a compact, user-friendly solution for continuous and dynamic aspiration monitoring in both home and institutional settings, holding promise for improving elderly healthcare.
Zhang et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: