1. Introduction
The measurement of physiological metrics, such as pulse rate and oxygen saturation, is of great interest to research, clinical, and commercial communities. Recent technological advancements have augmented our ability to access and evaluate physiological parameters related to human health and performance unlike ever before [
1]. Technologies for accurately monitoring pulse rate (PR) and oxygen saturation (SpO
2) are especially sought after for numerous applications, such as the ability to evaluate general health, and changes in physical and cognitive performance [
2,
3]. One of the most common techniques for evaluating measurements of blood oxygen saturation and cardiovascular performance is pulse oximetry by photoplethysmography (PPG) [
4,
5,
6]. Photoplethysmography is a simple, non-invasive, low cost, and portable optical technique that can measure pulse rate, oxygen saturation, blood pressure, and cardiac output, and it is generally accepted as providing valuable information about the general functioning, as well as acute changes, of the cardiovascular system [
2,
7].
A typical PPG device utilizes a light source and photodetector to measure changes within subcutaneous microvasculature to measure physiological data on the cardiovascular system from sites on the skin’s surface. The differential absorption of two wavelengths in pulsatile blood flow by oxygenated and deoxygenated hemoglobin allows for an accurate estimation of arterial oxygen saturation [
8,
9]. There are two general types of PPG techniques commonly used today, transmission and reflectance. Transmission PPG sensors detect light passed through the tissue and are therefore commonly used on peripheral sites, most commonly fingers and earlobes. Reflectance PPG sensors utilize a light detector adjacent to the emitter and are ideal for single-point contact readings [
10]. The most widely utilized locations for reflectance PPG includes the wrist, forearm, ankle, forehead, and torso [
11,
12]. Typically, these sensors are positioned on the skin using cuffs, or clips, as there is a required amount of pressure needed to apply the sensor to obtain the most accurate and reliable signal [
4,
13].
While PPG has been studied and used at various anatomical locations with varied applications, there are a number of limitations to the accuracy and effectiveness of these techniques, especially in dynamic or hypoxic environments [
6,
14]. It is known that the location of the sensor, high motion artifact, and issues with implementation of skin contact may compromise the effectiveness of the PPG sensor measurements [
4,
12,
15]. Tissue alterations, caused by both voluntary and involuntary movements, such as muscle contraction or dilation of tissues, can disrupt sensor readings, as physical displacement of the sensor from its original location can modify the PPG signal by changing the path of light [
16,
17]. Other factors influencing PPG reading accuracy include other anatomical or physiological differences such as skin color and amount of fluid retained by the tissues [
15,
18,
19].
Because of these various limitations of different PPG sensors at various anatomical locations, many studies have been conducted exploring their accuracy under different conditions. Other studies have collectively compared PPG at the finger, forearm, earlobe, ear canal, wrist, shoulder, forehead, chest, temple, neck, rib cage, wrist, lower back, and tibia [
12,
20,
21,
22,
23,
24,
25]. The consensus was that the forehead and finger locations provide the most accurate PR and SpO
2 measurements in static conditions, although both were found to be susceptible to deleterious effects from varying dynamic conditions, especially movement and desaturation events. While in a number of cases, the finger location was found to produce the most accurate results, Pertzov et al. determined that a centrally located PPG sensor, at the earlobe in their case, is more accurate than at the extremities like the finger and allows for earlier identification, treatment, and resolution of desaturation events [
26]. PPG sensor readings at the finger were found to be especially disrupted by movement. A study by Longmore et al. revealed that, among many anatomical locations tested, the forehead was the only location to record heart rate and SpO
2 within acceptable values both at rest and while walking, while the finger location readings were highly disrupted by movements [
12]. Conversely, in another study by Ross et al. comparing PPG sensors at high altitudes at the finger, forehead, and ear lobe, gold standard arterial blood gas results revealed the finger location to perform best, followed by the earlobe location and finally the forehead with the lowest performance for detecting hypoxia [
23]. Studies investigating the in-ear location for PPG sensors have found this location to be sufficiently accurate for detecting heart rate and oxygen saturation levels relative to the finger location and advantageous as a centrally located PPG sensor site [
21,
25,
27]. However, a study by Passler et al. determined that motion artifact due to normal jaw movements from chewing gum and talking throughout data recording had a significant deleterious influence on the in-ear location, resulting in a signal interference too intense to determine a precise pattern of pulse rate [
28].
While the finger and forehead sensor locations were found to be the most reliable and are the standard for validation of novel sensors, we can see they have some limitations. PPG readings at peripheral sites, like the fingers, are especially negatively influenced by motion artifacts, and other physiological factors, including blood pooling, blood pressure, and altered physiological responses [
29,
30]. PPG signals from the forehead are less affected by vasoconstriction and blood pooling compared to fingers [
31,
32]. However, there is a strong venous component to the forehead, as well as larger size of arteries and veins, which deteriorates the quality of the signal and increases unreliability of the measurements [
33]. Problems with motion artifact also negatively impact the accuracy and reliability of PPG readings at the forehead, as muscles in the forehead can create regular movement of the sensor over the skin, disrupting measurements [
34,
35].
Based on these known limitations of current PPG sensor locations, we determined that conditions of ideal PPG sites are ease of accessibility, good microvasculature, without large veins or arteries nearby, no hair or follicles, and low potential for motion artifact, in part affected by musculature, but also by location on the body. For example, the head typically moves less than the extremities and, as such, is a more ideal location for accuracy and reliability of readings in most dynamic circumstances. The location should also be centrally located and most directly in line with the heart to the brain to detect critical desaturation events as quickly as possible.
To meet these criteria and overcome the challenges of capturing accurate and reliable measurements, especially in critical or dynamic environments, we selected a largely unexplored location behind the ear, on the mastoid process, between the pinna and the hairline, above the sternocleidomastoid muscle, and near the posterior auricular artery for our SPYDR
® sensor suite (
Figure 1a). This mounted behind the ear location is most directly in line from the heart to the brain, is easily accessible, has little or no subcutaneous musculature or ligaments, is highly vascularized with no large arteries or veins, and is not in the extremities, subjected to higher motion and other location-based physiological alterations. Although, this location was previously demonstrated to be ideal mechanically and physiologically by He et al., it still remains largely unexplored and underutilized as a site for a reliable wearable vital signs monitor [
36]. The sensor in their study was found to experience deleterious effects of motion artifacts, but we believe this is related to the design on the sensor, not the location.
In order to further optimize PPG measurements at this location behind the ear, we developed an earcup-mounted device with a reflectance PPG sensor embedded within the ear-seal shown in
Figure 1b,c, called SPYDR
® (Standalone Performance Yielding Deliberate Risk). SPYDR employs an independently manufactured medical grade FDA-approved reflectance PPG sensor (OEMIII with 8000R reflectance sensor, NONIN
®, Plymouth, MN, USA) that is multi-wavelength (660 nm and 980 nm as pre-calibrated by the manufacturer) within the ear-seal of ear-muff type hearing protection devices like those found in aviation headsets and helmets (
Figure 1c). While embedded in the ear-seal, the sensor is up against the skin, with nothing else between the sensor and the skin. Soft black padding surrounds the sensor to prevent light leakage and maintain comfort of wearer as the sensor is held against the skin. This form factor alleviates issues with sensor application and contact with the skin. We custom fabricated the ear-seals to comfortably house the sensor such that it is held firmly against the wearer’s head behind the ear, in the specific location described above, in a self-contained unit housing the PPG sensor, battery, signal processing unit, and solid-state data storage. The SPYDR device in this study was installed within a military flyer’s helmet, such that it aligns with the helmet’s primary rotational axis and cranial support points, further reducing motion artifacts, but the sensor suite has potential for other applications as well.
The aim of this study was to determine if pulse rate (PR) and blood oxygen saturation (SpO2) could be accurately monitored using SPYDR installed in military flyer’s helmets. The gold standard arterial blood gas (ABG) PPG sensor validation analysis was conducted for SPYDR, along with two different commercially available finger sensors, to validate SPYDR for accuracy in measurements of PR and SpO2. In addition to accuracy against ABG, we also evaluated the SPYDR PPG sensor for continuity, and responsiveness for collecting data in hypoxic conditions in three simulated high-altitude, reduced oxygen environments to determine if it could produce reliable and accurate data for PR and SpO2 compared to two other commercially available PPG sensors at the finger and forehead. These tests conducted at varying oxygen levels, including moderate and severe hypoxic conditions, analogous to high altitude environments, were conducted using a reduced oxygen breathing device (ROBD). In this study, we determined we could accurately and reliably capture pulse rate and rapid changes in blood oxygen saturation behind the ear with SPYDR, with a faster response rate for changes in SpO2 compared to the PPG sensors at the finger and forehead.
4. Discussion
Previous research has discovered a number of factors that typically affect the accuracy of PPG readings in general, including the anatomical location of measurement, implementation of skin contact, application force, motion artifact, ambient temperature, and anatomy of the subject [
3,
4,
15,
39]. The objective of this research to meet these challenges by validating the accuracy and reliability of a novel device SPYDR, a PPG sensor suite, and its location behind the ear for measuring oxygen saturation (SpO
2) and pulse rate (PR), by comparing the device’s responsiveness to two commonly-used, commercially available, FDA approved PPG sensors at the finger and forehead.
We believe a number of benefits of SPYDR likely contribute to its accurate and reliable readings. This behind the ear location is optimal because of lack of hair follicles in the area, in both men and women, as opposed to an area in front of the ear where the readings may be encumbered by facial hair or a low hairline. Also, the behind the ear location is well suited for accuracy and reliability of readings because of low musculature in the area precluding excess movement, there are no major veins or arteries, and low movement of the cranium as opposed to the appendages, as the head typically moves less than the extremities. Motion artifacts, one of the largest limiting factors in PPG measurement accuracy, are usually caused by movement of the PPG sensor over tissue, skin deformation, or blood flow [
35,
40]. With SPYDR, we overcame many of the limitations of motion artifacts by location and design for the application of the sensor to the skin with the ear-seal design, which additionally blocks out disruptive ambient light.
As a first step in evaluating and validating the accuracy of the device, SPYDR was tested against simultaneous arterial blood gas (ABG) readings, the gold standard for novel PPG sensor validation. ABG readings were conducted by the highly renowned independent, university-based testing facility, the Hypoxia Research Laboratory of the University of California San Francisco, with testing protocols designed to generate data suitable for submission to the US Food and Drug Administration needed for novel PPG device approval. In these tests, continuous airway gas analysis by mass spectrometer is utilized, and pulse oximeter readings with SPYDR were compared to the direct arterial blood gas levels with multiwavelength oximetry. The root mean square (RMS) value takes into account both mean and standard deviation on the calculation, with a value of <3.5% considered accurate for reflectance PPG sensors. As shown in the results, SPYDR meets the standard criterion for PPG sensor accuracy with an RMS value of 2.61 for a hemioximeter range of 70–100%, as seen in
Table 2 and
Table 3.
ROBD tests were conducted to further explore the potential and validate the accuracy of SPYDR by comparing it to PPG sensor readings at the finger and forehead under conditions that cause dynamic changes to blood oxygen levels. In this test, SPYDR detected changes in oxygen saturation levels, while closely tracking a consistent pulse rate in hypoxic conditions faster than both sensors, and up to 50 s faster than the finger location (
Figure 6 and
Figure 7), as would be expected due to its location at the extremities. We also saw a general trend of a faster response time to rapid changes in SpO
2 levels by SPYDR compared to the forehead sensor, although the standard deviation was too high to consider the differences to be statistically significant. However, we believe the inability to detect a statistically significant faster response of SPYDR over the forehead location is due to the small test subject size, as in our own unpublished studies and observations we consistently find SPYDR to have the fastest response time behind the ear compared to any other site.
These factors combined suggest that monitoring pulse rate and arterial oxygen saturation with PPG behind the ear might circumvent the numerous effects of peripheral vasoconstriction and decrease the time to detect a hypoxic event. We were able to successfully validate the accurate performance of SPYDR in collecting pulse oximetry data at hypoxic levels of SpO2, while also monitoring the speed at which the device is able to detect a return to 100% SpO2 levels upon administration of 100% oxygen to subjects following hypoxic symptoms. Test subjects reported high satisfaction with SPYDR, reporting zero comfort or interference issues. As observed in previous tests, SPYDR earcups left uniform imprints on the user’s heads around the ears, allowing for identification of sensor placement after flight, but which did not cause pain of discomfort to the subjects.
Human cognition and survival are critically dependent on the availability of oxygen to the brain. However, the brain has no oxygen storage capacity of its own and therefore requires a continuous supply. Lack of brain tissue oxygenation can lead to a loss of nerve cell function in a matter of seconds. Previous studies have shown that a decline in core arterial oxygen saturation is delayed if the measurement is performed peripherally compared to a central location [
1,
6,
7,
8,
9]. A key feature of SPYDR’s behind the ear location is its closer proximity to the brain in line directly from the heart, as opposed to an extremity-mounted sensor, creating a more relevant analogue for cerebral oxygenation. It is ideal for determining levels of oxygen going directly to the brain, as oxygenated blood reaches the brain (via the carotid artery) faster than the extremities (capillaries of the finger) [
41]. The value of this location was evidenced by SPYDR’s ability to detect changing SpO
2 measurements prior to the observance of hypoxia symptoms reported in the subject, significantly faster than the other sensors at finger and forehead. Based on these anatomical and physiological advantages, we predict faster measurements with SPYDR at the onset of hypoxic conditions compared to sensors at other sites as well. We also believe our location is advantageous to other locations on the ear, such as in the ear-canal and earlobe because the ear-canal is highly subjected to motion artifacts from chewing and speaking, and the earlobe is consider an extremity that is susceptible to inaccuracies when peripheral perfusion is compromised, performing with less accuracy than a finger sensor in hypoxic conditions, as shown by Ross et al. [
23,
28,
42]
The faster detection rate of SPYDR to changes in blood oxygen saturation, measuring behind the ear, as compared to the finger or forehead, could provide earliest detection of hypoxic episodes. The development of alerting algorithms based on these data streams could likely provide users with key information and pre-emptive warning of impending emergencies, which have historically led to loss of life and destruction of property. This early detection capability could then be connected to an early warning signal, as is an ultimate goal of this device. Since these hypoxia experiments were performed on healthy volunteers, we are so far not able to make conclusions on the clinical usefulness of this technique. However, this was also outside the scope of the current study. Before a possible introduction in clinical practice, further investigation of the technique under more dynamic conditions should be considered.
As the aviation community attempts to maximize training effectiveness and human performance while mitigating risk, an objective tool to quantify cognitive status and draw attention to degraded cognitive performance is needed. To date there are no reliable devices within tactical aviation to monitor oxygen levels of fighter pilots. This is critical because it is hypoxia and loss of consciousness of the pilot, not equipment failure, that leads to most fatal crashes. Under periods of high acceleration, the body’s cardiac output is placed under considerable strain. The heart’s efforts to circulate oxygen rich blood to the brain are countered by the aircraft’s centripetal acceleration, causing reduced oxygenation of the brain and, in severe cases, an ischemic loss of consciousness. SPYDR represents an important opportunity to collect data for a relative performance index and could ultimately improve current risk mitigation techniques, as well as better detect and alert pilots to a degraded cognitive status due to hypoxia. Ease of application and implementation extend the quantity of applications and industries for SPYDR. Future studies exploring its accuracy in highly dynamic environments associated with tactical flight would provide further valuable information for validating the feasibility of SPYDR in the field. Additionally, with continual advances in data science and machine learning, expanded research on in-flight data derived from SPYDR may result in enhanced biometric quantification of human health, performance, and cognitive status in addition to early warning signals for hypoxia.
The standard FDA testing protocol for novel PPG sensor suites requires that all subjects be healthy and free of preexisting conditions, defined primarily as being nonsmokers with no evidence of lung disease, obesity, or cardiovascular problems, that would negatively impact arterial blood gas results and put the subjects at risk. Therefore, one limitation of this study is that all subjects were classified as “healthy” and the results here do not show if there will be similar readings in subjects with preexisting health problems, such as lung disease or cardiovascular illness. Also, while some applications within this study are not those typically seen in patients, such as simulated high altitudes, the hypoxic conditions of reduced blood oxygen tested under these criteria can be analogous to physiological responses seen in other pathological or acute adverse conditions. We therefore would expect the SPYDR sensor suite at this specific behind the ear location to still perform reliably in patients with preexisting health conditions in future studies, creating great potential for this technique in the medical field as well.
The data presented in this preliminary study suggest that the SPYDR sensor suite reading behind the ear provides accurate measurements for physiological parameters of cardiovascular performance that is as accurate and reliable as finger and forehead PPG sensors with a faster response to detecting changes in blood oxygen saturation levels.