An accelerometry and gyroscopy-based system for detecting swallowing and coughing events
Stevens, Guylian,Van De Velde, Stijn,Larmuseau, Michiel,Poelaert, Jan,Van Damme, Annelies,Verdonck, Pascal
DOI: https://doi.org/10.1007/s10877-024-01222-6
IF: 1.977
2024-09-22
Journal of Clinical Monitoring and Computing
Abstract:Measuring spontaneous swallowing frequencies (SSF), coughing frequencies (CF), and the temporal relationships between swallowing and coughing in patients could provide valuable clinical insights into swallowing function, dysphagia, and the risk of pneumonia development. Medical technology with these capabilities has potential applications in hospital settings. In the management of intensive care unit (ICU) patients, monitoring SSF and CF could contribute to predictive models for successful weaning from ventilatory support, extubation, or tracheal decannulation. Furthermore, the early prediction of pneumonia in hospitalized patients or home care residents could offer additional diagnostic value over current practices. However, existing technologies for measuring SSF and CF, such as electromyography and acoustic sensors, are often complex and challenging to implement in real-world settings. Therefore, there is a need for a simple, flexible, and robust method for these measurements. The primary objective of this study was to develop a system that is both low in complexity and sufficiently flexible to allow for wide clinical applicability. To construct this model, we recruited forty healthy volunteers. Each participant was equipped with two medical-grade sensors (Movesense MD), one attached to the cricoid cartilage and the other positioned in the epigastric region. Both sensors recorded tri-axial accelerometry and gyroscopic movements. Participants were instructed to perform various conscious actions on cue, including swallowing, talking, throat clearing, and coughing. The recorded signals were then processed to create a model capable of accurately identifying conscious swallowing and coughing, while effectively discriminating against other confounding actions. Training of the algorithm resulted in a model with a sensitivity of 70% (14/20), a specificity of 71% (20/28), and a precision of 66.7% (14/21) for the detection of swallowing and, a sensitivity of 100% (20/20), a specificity of 83.3% (25/30), and a precision of 80% (20/25) for the detection of coughing. SSF, CF and the temporal relationship between swallowing and coughing are parameters that could have value as predictive tools for diagnosis and therapeutic guidance. Based on 2 tri-axial accelerometry and gyroscopic sensors, a model was developed with an acceptable sensitivity and precision for the detection of swallowing and coughing movements. Also due to simplicity and robustness of the set-up, the model is promising for further scientific research in a wide range of clinical indications.
anesthesiology