Investigating and developing low-cost wearable respiration sensors.

Abstract

Respiration is a vital parameter in healthcare monitoring, in which it can be used to identify and help prevent illnesses such as sleep apnoea and sepsis. Small changes in respiration can be linked to deteriorating health, where failing to detect these changes promptly can often result in poorer outcomes for the patient, and in serious cases mortality. In some cases where respiratory rate is ≥ 27 BPM, it can be a better indicator for cardiac arrest from up to 72 hours. This brings the importance of continuous respiration monitoring of time greater than 1 hour. The current sensors that are available for use can monitor respiration rate continuously, but not all of them can for long periods of time. An example on the importance of continuous monitoring is sepsis, where continuous monitoring could be used to identify early risk warnings for the medical professionals. Many sensor technologies showed potential in detecting respiration, but the capaciflector sensor shined due to the low number of research done on it for applications in respiration detection, and due to the ease of use. The sensor is developed into a sensor system that is manufactured and compared to a pneumotachometer and a belt sensor which are both gold standard sensors in respiration monitoring. The data is directly compared using Bland Altman Statistical Analysis, with limits of agreement being ±3 BPM for the difference between the capaciflector and the gold standard sensor. This research aims to identify and develop a low cost wearable respiration sensor that is capable of accurately (within ±3 BPM) measuring respiration, as well as continuously monitoring respiration over long periods of time (≥ 1 hour). Two short studies with 70 participants were conducted to assess the capaciflector as a respiration sensor, where study one was done on stationary participants and the second study on participants riding a bicycle. A new algorithm was developed to process and analyse the data, which resulted in biases and lower limits of agreements of 0.05 ± 0.04 BPM and 0.70 ± 0.20 BPM in the stationary tests respectively. While the biases and agreements are 0.12 ± 0.03 BPM and 3.02 ± 0.13 BPM for the bicycle tests respectively. A new algorithm was developed and tested in MATLAB based off a short Fourier transform which analysed the signal in the spectral plane. The results from these studies allowed a capaciflector which is PCB based to be developed. The hardware was developed to match the needed specifications and was tested with the same algorithm from the studies. The new hardware contained better and improved methods for reading the capacitance from the capaciflector, while also having more sensors such as a real time clock and an accelerometer. Results from testing the hardware showed potential in utilising the accelerometer data as part of the processing algorithm. The hardware developed is capable of continuously monitoring respiratory rate for periods spanning more than 48 hours on a single CR2032 coin cell battery. Thus by running a small trial of 10 participants, the new sensor system was tested alongside a newly developed algorithm that can segment stationary data from data that contains movement artefacts. Three studies were conducted, the first study is a metronome study which resulted in a bias of 0.04 BPM with ±0.54 BPM for the limits of agreement across all 10 participants. While the second study which is a walking study achieved results of -0.04 BPM for the bias and ±1.48 BPM for the limits of agreement. This study utilised the new algorithm which allowed the results to be within the required accuracy even while including walking data. And when compared against the initial results found, the algorithm developed proved to be very useful for monitoring respiration rate. Finally the last study is a study that aimed to explore the effects of speaking on respiratory data, in which the study showed that speaking has a unique pattern of being similar to a saw tooth wave which is reflected in the respiratory data. Overall the capaciflector based sensor system developed proved to be within the required aims as well as satisfying the conditions set, with being compact, comfortable and capable of continuously monitoring respiration rate over long periods of time. And this research demonstrates that it is viable for use towards long term home monitoring or ambulatory care patients.

More information

Identifiers

ORCID for Mahdi Mohamed Saleh Abdulla Ahmed Shaban: orcid.org/0000-0002-7219-3708

ORCID for Neil White: orcid.org/0000-0003-1532-6452

Catalogue record

Export record