Development and evaluation of thermal imaging techniques for non-contact respiration monitoring

Respiration rate is one of the main indicators of an individual's health and therefore it requires accurate quantification. Its value can be used to predict life threatening conditions such as the child death syndrome and heart attacks. The current respiration rate monitoring methods are contact based, i.e. a sensing device needs to be attached to the person's body. Physically constraining infants and young children by a sensing device can be stressful to the individuals which in turn affects their respiration rate. Therefore, measuring respiration rate in a non-contact manner (i.e. without attaching the sensing device to the subject) has distinct benefits. Currently there is not any non-contact respiration rate monitoring available for use in medical field. The aim of this study was to investigate thermal imaging as a means for non-contact respiration rate monitoring. Thermal imaging is safe and easy to deploy. Twenty children were enrolled for the study at Sheffield Children Hospital; the children were from 6 month to 17 years old. They slept comfortably in a bed during the recordings. A high resolution high sensitivity (0.08 degree Kelvin) thermal camera (Air A40) was used for the recordings. The image capture rate was 50 frames per second and its recording duration per subject was two minutes (i.e. 6000 image frames) A median digital lowpass filter was used to remove unwanted frequency spectrum of the images. An important issue was to localize and track the area centered on the tip of the nose (i.e. respiration region of interest, RO!). A number of approaches were developed for this purpose. The most effective approach was to identify use the warmest facial point (i.e. the point where the bridge of the nose meets the corner of one of the eyes). A novel method to analyse the selected ROI was devised. This involved segmenting the ROI into eight equal segments centred on the tip of nose. A respiration signal was produced for each segment across the 6000 recorded 'images from each subject. The study demonstrated that the process of dividing the ROI into eight segments improves determination of respiration rate. The respiration signals were processed both in the time and frequency domains to determine respiration rates for the 20 subjects included in the study. The respiration values obtained from the two domains were close. During each recording respiration rate was monitored using conventional contact methods (e.g. nostril thermistor, abdomen and chest movement sensor etc). There was a close correlation (correlation value 0.99) between respiration values obtained by thermal imaging and those obtained using conventional contact method. The novel aspects of the study relate to the development of techniques that facilitated thermal imaging as an effective non-contact respiration rate monitoring in both normal and patient subject groups.EThOS - Electronic Theses Online ServiceGBUnited Kingdo