A PCA-based method to select the number and the body location of piezoresistive sensors in a wearable system for respiratory monitoring
Regular monitoring of several physiological activities like breathing, blood pressure, heart functioning and body temperature are required for monitoring health and various other applications. Monitoring Respiratory Rate (RR), which is sensitive to environmental and physiological stressors, is essential to measure the patients' actual status. The researchers have devised a method to select the number of piezoresistive sensors and their position on the body to instrument wearable systems for estimating RR at rest and during walking/running tasks.
Respiratory Rate can be monitored by: contact-based and contactless techniques. Among the various approaches to designing wearable systems, the piezoresistive sensors support the sensor structure and could be used as the support materials and sensitive element. The compact and highly integrated flexible systems are comfortable, wearable and adaptable to be integrated, sewed, or embedded within clothes. The piezoresistive sensors in smart-textile for RR monitoring also has an advantage for the simple electronics needed to work them and low power consumption, making it an affordable solution in unstructured environments.
The approach is based on Principal Component Analysis (PCA), which address the concerns for the optimal number of sensors to be used and their appropriate body location, both at rest and during physical activity. It is used to determine the most suitable sensors to interpret the recorded information accurately. In other words, only those sensors that showed weight on estimated Principal Components higher than 15% were retained. In addition, to further remove as much redundant information as possible, redundant sensors were checked by implementing a linear correlation between the retained ones. If the correlation coefficient between two sensors was higher than 0.8, the one with a smaller weight on Principal Components was discarded.
Breathing-unrelated movements (e.g., caused by activities of daily living) strongly affect the quality of the signals recorded by piezoresistive sensors. The smart garment instrumented by six piezoresistive-based smart textile sensors positioned on both sides of the lower thorax, upper thorax and abdomen is developed. It was tested by PCA-based method by performing six trials on a treadmill: one trial at rest (absence of breathing-unrelated movements), four walking trials (speed of 1.6 km·h-1, 3.0 km·h-1, 5.0 km·h-1 and 6.6 km·h-1) and a low-speed running trial at 8.0 km/h. During each trial, the RR was monitored using both the proposed smart garment and a flowmeter, which was used as a gold-standard system to understand the wearable performance.
The inference was based on two types of analyses. One was related to implementing the sensor selection method, thus mainly devoted to determining the sensors to retain. The other one was to assess whether that specific sensor combination provides an accurate estimation of the RR compared to the reference flowmeter in terms of both average and breath-by-breath RR.
The experimental results showed that: breathing assessment at rest requires one sensor placed on the lower thorax. Low speed walking assessment requires three sensors placed on the upper thorax, lower thorax and abdomen. High speed walking and running assessment requires a four-sensor configuration (one sensor placed on the upper thorax, one on the lower thorax and two on the abdomen).
The proposed analysis and sensing technology may pave the way for future study and investigations where breathing biomechanics could be pivotal in determining patients' health regarding respiratory functions and help in rehabilitation.