1. Introduction
Photoplethysmography (PPG) is an optical measurement technique used primarily for detecting volumetric changes of pulsatile blood flow in vascular tissue. It takes advantage of the fact that light is absorbed by body tissues, such as skin, fat, bone, and blood, in different amounts for any given wavelength of light. During the systolic phase of the heart, when blood is pumped from the pulmonary circulation to the systemic circulation via the left atrium and left ventricle of the heart, there is a momentary increase in the volume of blood in arteries, and the arteries expand. This increase in volume causes more light to be absorbed, resulting in less light being transmitted through the tissue. Utilising suitable sensors and instrumentation, these changes in light intensity can be detected and recorded, and the “pulses” detected can be used for anything from simple heart rate calculations to assessing cardiovascular health [
1].
There are two modes of sensors used in PPG, transmission and reflectance. In transmission mode, the light source and detector are placed opposite one another inside a finger clip, and light is transmitted through the finger tissue. This is the most common mode seen mainly in pulse oximeters. The second type of mode is reflectance, where the source and detector are placed next to each other, and light is backscattered within the tissue. This mode is most commonly found in wearable devices, watches and armbands, or incorporated into mobile technology. Irrespective of the type of sensor used, the acquired PPG signal exhibits a quasi-periodic pattern consisting of an arterial pulse wave for each heartbeat. The morphology of this arterial pulse is known to be influenced by several physiological variations such as heart rate, heart rhythm, stroke volume, arterial stiffness, blood pressure, respiration, and the autonomic nervous system. This responsivity of the PPG signal to various physiological processes has helped researchers derive various diagnostic markers for vascular ageing and arterial compliance, large arterial stiffness, hypertension risk stratification, total peripheral resistance, atrial fibrillation, stress, endothelial dysfunction, gingivitis, apnoea, and autonomic nervous system (ANS) responses [
2].
However, various other factors that are not physiological, such as sensor geometry, skin-sensor interface, contact pressure exerted by the sensor clip, photodiode sensitivity, ambient light, and measurement site, can also exhibit morphological changes in the PPG signal. In particular, the contact force or contact pressure (CP), the pressure exerted by a PPG sensor on the measurement location, is thought to influence the quality and morphology of the PPG signal significantly. Simple morphological features such as the systolic peak (used to calculate the heart rate), pulse width, pulse area, and the relative AC/DC amplitudes of red and infrared signals (used for oxygen saturation calculation) are thought to be adversely affected by an increase in sensor contract pressure [
3]. In the past, several researchers have tried to investigate the effect of sensor contact pressure on the quality of the PPG signal. A summary of their findings, along with the type of sensor used and the type of force applied, is presented in
Table 1. To make a likewise comparison between the studies and to underpin their findings, the contact pressure as described in the papers presented in
Table 1, column 3 were converted to millimetres of mercury–mmHg and reported in
Table 1, column 4. Every effort has been made to correctly identify the experimental setups of the reviewed studies so as to report the correct equivalent contact pressure. The force to pressure conversion was calculated using Equation (1):
As can be seen from the table, the effect of CP on the PPG signal acquired or the parameters estimated from the PPG signal is significant. Following these research paths still leaves a number of research questions viable for investigation. These are:
How much is the degree of change in PPG signal features that an increase in sensor CP can create?
What are the PPG signal features that are most affected?
Is the change in a particular PPG signal feature significant enough to create misinterpretations when deriving diagnostic indices?
These questions and the question of the potential behaviour of arteries located under the PPG sensor when applied to wearable devices has not been investigated [
2]. Other factors, such as the size of measurement location (e.g., finger diameter), have never been considered in any of the studies. Moreover, the in vivo studies reported thus far fail to isolate the sensor contact pressure at known blood pressure states specifically. This is obviously mostly due to the nature of recruiting volunteers or patients (who would naturally all vary widely in blood pressures due to physiological differences) and the known difficulty with trying to induce specific blood pressure states. Although methods such as the Valsalva Manoeuvre (VM) [
10,
11] have been used to induce lower pressures temporarily, this cannot be maintained indefinitely and is difficult to repeat successfully and concurrently to enable repeated measurements.
All the above factors have led to a growing consensus that external factors that can affect the PPG signal quality and morphology, such as contact pressure, need to be further studied rigorously to help standardise PPG sensor design, which will, in turn, facilitate ideal PPG signal acquisition. This will also help standardise PPG measurements so that they are useful in not only standard monitoring (pulse oximetry and heart rate) but also for future and emerging applications such as Pulse Rate Variability (PRV) or Pulse Wave Velocity (PWV) measurements to assess cardiovascular health.
As a first step towards discovering the optimal PPG sensor design and conditions, an in vitro investigation was carried out using a relatively simple vessel-tissue phantom with similar mechanical properties to human anatomy. The developed in vitro vessel-tissue phantom with a pulsatile fluidic flow was able to maintain and replicate various blood pressure states whilst under controlled and measurable compressive loads, and was simultaneously able to record PPG signals directly at the site of compression. Through this effort, we aim to determine what sensor-contact pressure is optimum to obtain PPG signals and explore which morphological PPG features are least or most affected by increasing contact pressure.
4. Discussion
The quality and morphology of a PPG signal are known to be affected by several external factors. One such factor that is known to play a significant role is the sensor contact force or pressure. Several researchers including Teng and Zhang [
4]; Grabovskis et al. [
5]; Shimazaki et al. [
6]; Kasbekar and Mendelson [
7]; Lee et al. [
8]; and Scardulla et al. [
9] have previously tried to investigate the effect of contact pressure on the quality of the PPG signal and the basic morphological changes such as pulse amplitude. With more and more researchers now investigating the use of photoplethysmography (<550 peer-reviewed publications in 2020) for measuring parameters ranging from PRV to biometric recognition, it is important to understand the effect of sensor contact pressure on the morphological features of the PPG signal, particularly since very many of these measurement methods are based on detecting appropriate changes in the morphological features of the signal.
In an attempt to understand the above, in this work, we investigated the effect of sensor contact pressure on an in vitro vessel-tissue phantom which mimics the properties of human tissue. The phantom was built such that the PPG sensor can be placed directly (3 mm deep) above the vessel, facilitating the investigation of our research questions.
The nature of in vitro investigation, however, is limiting in that some interesting physiological phenomena and anatomical features are difficult or, at present, impossible to replicate. In favour of the simplified model presented, however, it does allow the investigation of simple external stimuli, such as contact pressure, without complicating the response with factors outside our control, such as vasodilation/contraction or age-related vessel compliance. The results of the in vitro experiment can therefore be used to further inform or compare results from future in vivo studies.
4.1. SNR Analysis
It has been shown in previous studies that there exists optimum sensor contact force/pressure and the reported physiological reading when compared to another or gold standard recordings [
4,
5,
6,
7,
8,
9]. In our experiment, the reported SNR values across the measurements show little variance among either the repeated experiments or between the blood pressure states, and an optimum sensor pressure is observed that is similar to the previous in vivo studies. This lends confidence to the in vitro method we have described, and with further development of our phantoms, a larger set of physiological parameters may be tested. It must be noted, however, that in the studies by Kasbekar and Mendelson, Shimazaki et al. and Scardulla et al. [
6,
7,
9], an emphasis was mostly put on reducing motion artefact during exercise, whereas the study by Lee et al. [
8], focused on the subtle respiratory variations on the PPG. Both of these scenarios (physical exercise and respiratory artefacts) were not simulated in our study.
The study by Grabvoskis et al. [
5] was of particular interest to us, as the effects of CP on PPG morphology to assess arterial stiffness is an interesting development in the field of PPG for other physiological monitoring other than HR and SpO2, i.e., vessel compliance or arterial function monitoring. However, the study protocol did not allow us to make the same type of analysis as in Grabvoskis’ study, and further experiments would be needed to make a comparative analysis.
Whilst it is observed that the red PPG signals have lower SNR generally than the infrared signals, this can be explained by the fact that the infrared PPG monitoring in this phantom may be more robust either due to the sensor component geometry, the type of ink used or a combination of the two. It is also interesting to note that this relationship seems to reverse during the stage 2 hypertensive experiments and might be explained by some mechanical property of the phantom, though further analysis, and more rigorous experiments and a review of the phantom construction may be needed to confirm this. To test the hypothesis that this is related to a mechanical property, a new in vivo experimentation protocol should be developed to investigate whether the PPG signal can be used to directly infer a mechanical condition of anatomy, such as vessel compliance.
4.2. PPG Feature Analysis
It has been shown that the morphological features of the PPG are affected by the contact force induced by the sensor to the tissue [
4,
5,
9]. The obtained results in this in vitro study show that the contact pressure affects primarily amplitude-related features, such as the area of the cycle or the pulse amplitude. On the contrary, time-based features, such as the duration of the cycles or the width of the cycle, are less affected by the contact force of the sensor, regardless of the blood pressure. Hence, applications that are mainly based on time-related features, such as pulse rate variability (PRV) analysis, could be performed regardless of the contact force applied by the sensor in the tissue, as long as the detection of the pulses does not rely on systolic peaks, as has been suggested in by Mejía-Mejía et al. [
18]. However, in applications where amplitude features (pulse amplitude, area of the systolic peak and diastolic peak, Y-coordinate of the centroid of the systolic and diastolic pulses) are key to deriving diagnostic markers while measuring oxygen saturation in pulse oximetry, assessing vasoconstriction, vasodilation, venous function, measuring the ankle pressure, genital responses, blood pressure, and cardiac function, contact pressure plays a major role [
19,
20,
21,
22,
23,
24]. Physiologically, the PPG amplitude features are a result of a complex interaction of stroke volume, vascular compliance, and tissue congestion effects [
25]. Hence non-optimal contact pressure could potentially result in incorrect assessments.
Interestingly, the variation among the repeated experiments for each measurement was less notorious for features extracted from red signals. This could have two different explanations; the red signals are less affected by noise in the in vitro setup and are more robust due to the type of dye used (India ink which is absorbed more by the red light than the infrared light in accordance with the complementary colour chart), or the features extracted from red PPG signals tend to be less affected by noise. Although it is not possible to certainly pinpoint the reason from this current study, further studies should aim to explain these differences, and they should take into account that the red features might be more repeatable in the face of different contact forces or in scenarios in which the contact force cannot be controlled.
Furthermore, the behaviour of an arterial blood vessel located directly underneath the PPG sensor can be derived from this investigation. Physiologically, arteries have two important functions [
26]. The first is a conduit function that allows blood to reach the periphery of the whole body; the second is a cushioning function that minimises sudden surges in the pressure of blood vessels by the stroke volume during a systolic cycle of the heart. For these purposes, arterial walls must be compliant enough to temporarily store a portion of the blood during systole and release it during diastole. Healthy arteries are highly distensible and show a nonlinear stress–strain response with an exponential stiffening effect at higher pressures [
27]. This stiffening effect, common to all biological tissues, is based on the recruitment of the collagen fibrils, which are extremely stiff [
28]. The arterial wall in the living body is pre-stretched, under a blood pressure load in the blood vessel; therefore, they are always in a stressed state. However, when an external pressure to the outer arterial wall is applied for any reason (such as when using a PPG sensor), the stress in the arterial wall decreases. When external pressure, for a particular case, is the same as the mean blood pressure in a vessel, the stress in the arterial wall becomes minimised. Since the exponential stiffening effect eventually disappears for this condition [
29], the maximum volumetric change of the arterial wall by blood pulsation can be monitored by PPG [
30]. This effect is perfectly exemplified by
Figure 4. As can be seen from
Figure 4, the amplitude of the PPG signal in
Figure 4 increases with increasing sensor contact pressure up to a point where the external CP is similar to the internal pumping pressure, where a maximum SNR value is recorded before the amplitude begins to decrease, as the vessel in the phantom is slowly becoming more restricted, until there is a sudden drop off in SNR when the vessel is completely occluded and the SNR becomes negative (signal noise is dominant). The same effect repeated over several blood pressures should result in a positive pressure correction between the PPG oscillometry peak and the blood pressure inside the vessel.