Use this URL to cite or link to this record in EThOS:
Title: Non-contact vital sign monitoring of pre-term infants
Author: Chaichulee, Sitthichok
ISNI:       0000 0004 8502 9012
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Non-contact vital sign monitoring using video cameras enables the measurement of heart rate and respiratory rate to be performed without sensors attached to the skin of the patient. This provides advantages in terms of patient comfort and the management of skin irritation and infection. However, the use of non-contact technologies in the clinic presents several challenges such as variations in human skin colour, changes in lighting conditions, the detection of the presence of a patient in the video frame and the selection of suitable regions of interests (ROIs) from which vital signs can be estimated. This thesis proposes a framework for the accurate and continuous measurement of heart rate and respiratory rate in pre-term infants in a real-world hospital environment. The framework has been developed and validated on the data obtained from a clinical study of pre-term infants hospitalised in the high-dependency area of the neonatal intensive care unit (NICU) of the John Radcliffe Hospital in Oxford. The study involved the recording of videos and reference vital signs for 90 sessions during daytime from 30 pre-term infants, comprising a total recording time of 426.6 hours. The dataset provided a wide range of vital sign values for developing and validating the framework. Multi-task deep learning algorithms were developed so that vital-sign estimation could be performed only when the infant was present in front of the video camera and no clinical interventions were undertaken. The algorithms are able to deal with different skin tones, lighting condition changes and variable body postures. Heart rate is estimated from the photoplethysmographic imaging (PPGi) signal derived from subtle changes in the colour of the skin areas. Respiratory rate is estimated using data fusion from the PPGi signals along with the shape and morphological properties of the skin areas. Signal quality assessment algorithms are developed for both heart rate and respiratory rate to discriminate between clinically acceptable and noisy signals. In a test set, the mean absolute error between the reference and camera-derived heart rates is 2.3 beats/min for over 76% of the time for which the reference and camera data are valid. The mean absolute error between the reference and camera-derived respiratory rate is 3.6 breaths/min for over 78% of the time. Most of the time periods during which the video camera cannot provide estimates of heart rate and respiratory rate are lower than 30 seconds. This thesis further investigated the feasibility of using non-contact algorithms to estimate heart rate and respiratory rate during a painful clinical procedure, a heel prick for withdrawing blood for blood gas analysis. Adaptation of the algorithms enables physiological indicators of pain to be derived from the analysis of the video camera data.
Supervisor: Tarassenko, Lionel ; Montoya, Mauricio Villarroel Sponsor: Royal Thai Government Scholarship ; Research Councils UK Digital Economy Programme
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Biomedical engineering