AbstractBreathing Rate (BR), which is an important indicator of deterioration, has been widely neglected in hospital settings due to the requirement of invasive procedures and skilled nurses for it to be measured. However, BR can be estimated using non-invasive techniques from Electrocardiograms (ECG), Photoplethysmograms (PPG) and Seismocardiograms (SCG) recordings. Nonetheless, the current state of the art non-invasive BR estimation methods have not been broadly evaluated, especially using real data from hospitalised patients, thus are still inaccurate and unreliable in real settings.
Another critical deterioration indicator, especially on paediatric and neonatal wards, is the Work-of-Breathing (WOB). The WOB indicator provides information about the effort expended by a patient while breathing, however its evaluation is largely subjective and until now it has only be assessed by skilled nurses. Regardless of the clinical significance of the WOB, its evaluation from ECG signals has received scant attention in the research literature.
This thesis presents investigations of the standard BR estimation methods in order to develop the theory and implementation know-how around BR extraction from ECG and SCG signals and address the limitations of the current methods with a view to benefit continuous patient monitoring and triage. Advanced signal processing techniques such as Empirical Mode Decomposition (EMD) have been exploited and enhanced, leading to a novel filter-based EMD algorithm for continuous BR monitoring which does not require the identification of ECG features.
In the final part of the thesis the WOB estimation from real ECG signals is explored. A measure based on the R-to-S (RS) amplitude variability due to respiration in the ECG signal was developed in this thesis and it was found for the first time that there is a positive monotonic relationship between the RS variability and the WOB.
All developed methods in this thesis are tested on real children data obtained from Birmingham's Children Hospital. The proposed BR methods in addition to the WOB measure demonstrated good estimation of their corresponding physical indicators on this real data.
|Date of Award||Jan 2020|
|Supervisor||Randa Herzallah (Supervisor)|
- patient monitoring
- signal processing