A systematic approach to improving the signal processing and illumination in imaging photoplethysmography

The utilisation of opto-electrical techniques in biomedical monitoring has become increasingly widespread, largely due to its non-invasive nature and the relatively low requirements for the hardware setup. Imaging photoplethysmography (iPPG) – a method of detecting blood volume changes in-vivo – has been a region of intensive research for the past decade focusing on validation and comparison with other microvascular measurement techniques. Although the outcomes from present research activities show promising results when tested in a controlled environment, they lack certainty when applied to the real-world scenarios, e.g. an operating theatre or a walk-in centre. To effectively access physiological information through iPPG, a number of key issues should still be addressed. There are four independent yet interconnected fields contributing to the successful implementation of iPPG as a clinical monitoring method. Opto-physiology, as the leading one, grasps the understanding of how light penetrates and interacts with biological tissues. Optics and light sensors are responsible for catching blood volume change in the tissue, therefore, careful selection of purpose-built hardware to work in challenging hospital conditions, such as in the presence of ambient and stray light, is crucial. Illumination in various wavelength spectrum has been shown to behave differently while interacting with human tissue and can reveal unique physiological patterns, which makes illumination another critical system to study. Lastly, motion-induced noise is a common unwanted companion in optical imaging techniques, thus making signal processing the fourth system to be targeted. One motivation for this study, amongst others, is the lack of a standardised framework for assessing how various aspects, including iPPG signal extraction and motion artefact reduction, or variability in light uniformity and intensity, could influence estimation of the heart rate and the construction of blood perfusion maps. The aim of this thesis is to investigate whether the reliability of iPPG measurements in real-world settings could be improved by tuning optical equipment and signal processing algorithms in a systematic and repeatable manner, achieved by the use of a human hand cast that does not exhibit temporary changes unlike live tissue. The contribution of this study is the detailed description of the individual elements of iPPG illumination sources and the experimental protocols implemented to determine their individual influence on the extracted iPPG signals, which has not been previously made in such detail. The utilisation of a hand cast shows how both the iPPG signal processing and the optical setup could be improved in a repeatable manner to deal better with motion artefacts, and how to develop a better understanding of whether the obtained results have been influenced by an individual subject or by a systematic error in iPPG systems.