Additively Manufactured High Gauge-Factor Compliant Strain Sensor for Machine Learning-Based Sleep Apnea Detection and Prediction

This article presents a methodology for fabricating a strain sensor designed explicitly for respiratory monitoring. The sensor’s high gauge factor (GF) and low stiffness make it particularly suitable for application to patients who cannot tolerate high stress on their chest, such as infants or the elderly. The sensor comprises a flexible polymeric spring coated with a carbon-based nanocomposite acting as the active layer. Different weight percentages of carbon-based materials are used to determine the percolation threshold, and the percolation threshold was determined to be at 4.75 wt% and 4.25 wt% of graphene and multi-walled carbon nanotubes (MWCNTs) with maximum GFs of 949.02 and 117.52, respectively. Scanning electron microscopy (SEM) and GF analysis showed that a higher loading would lead to a more brittle sensor and premature failure due to the contrast in mechanical properties between the active layer and the flexible substrate. A cyclic fatigue test was done on the strain sensor under normal operating conditions. The sensor was able to withstand a total of 21600 cycles without failure with some relatively consistent signal. Additionally, we study the effectiveness of utilizing respiratory signals for sleep apnea (SA) event prediction using short-time Fourier transforms (STFTs) and 2-D convolutional neural networks (CNNs), a suitable use case for fabricated sensors. The results provide a 1%–3% improvement in accuracy over traditional baselines that utilize raw signal data.