4.1.1. PPG Signals

Figure 5b shows PPG measurement, which hinges on the acquisition of the BVP waveform, by illuminating a tissue volume with an LED and measuring the transmitted light through the tissue with a PD (PDT) or the light reflected back from the tissue volume with a PD (PDR) [53,77]. The measurement is most commonly obtained from the finger in transmissive mode, but in reflective mode can be measured from other well perfused sites on the body (e.g., forehead, forearm, wrist). PPG is the basis for pulse oximeters, used ubiquitously for measuring arterial oxygen saturation in clinical settings. With each heartbeat, the volume of arterial blood in the tissue being illuminated decreases during diastole and increases during systole, and thus the light passing through the tissue is brighter (diastole) and dimmer (systole) at the photodiode. Since the volumetric expansion and contraction of the arteries is dependent on pulse pressure and arterial compliance, the PPG waveform closely resembles the underlying arterial blood pressure waveform in shape. While PPG waveforms are captured on commercially available pulse oximeter instrumentation, such waveforms may not be reliable for CRM since the PPG signals are heavily filtered and processed [77]. PPG can be measured in both transmissive and reflectance mode: for transmissive mode operation, the LED and photodiode are on opposite sides of the tissue (typically the earlobe, fingertip, or toe); for reflectance mode operation, the LED and photodiode are adjacent to one another on the same side of the tissue, and thus the locations for measurement can theoretically be anywhere on the body with sufficient perfusion (e.g., forehead, forearm, chest, and wrist). The main disadvantages of reflectance-mode PPG are that the signal quality is lower [78], the measurement varies with positioning and the distance between the LED and the photodiode, and the signal is more affected by motion artifacts [79,80]. Recent developments in device fabrication have allowed PPG sensing systems to be flexible and skin-interfaced for comfortable use in long-term care scenarios [81,82]. Soft and stretchable optoelectronics sensing for transmissive PPG measurement was demonstrated by Biswas, et al. [83]. An interesting approach not requiring an LED but rather using ambient light for PPG sensing was demonstrated by Han, et al.; with this approach, PPG signals with distinguishable heartbeat peaks were recorded and corresponding pulse oximetry readings were obtained [84].

#### 4.1.2. Tonometry Signals

Figure 5c shows tonometry measurement, which involves the application of a force to flatten (or applanate) an artery with a given applanation force (Fappl), and a pressure sensor applied to the skin above the artery then records the time varying fluctuations in pressure applied by the blood on the arterial wall [85,86]. With perfect applanation, this pressure waveform (BP(t)) is exactly equal to arterial pressure; however, in practice, applanation is usually imperfect and thus the waveform simply resembles BP. The most common measurement site is the radial artery. The advantage in tonometer measurements as compared to PPG is that substantially lower power consumption is required [87]—PPG employs active sensing where light is delivered to the tissue and then the resultant transmitted or reflected light level is detected; tonometry is a passive measurement where a transducer simply records the distension of the arterial wall. However, the major disadvantage in tonometry as compared to PPG is that the measurement is highly dependent on the location, and the transducer must be reliably placed over a superficial artery.

#### 4.1.3. Wearable Ultrasound

Figure 5d shows ultrasound array based measurement, which uses an array of ultrasound transducers in a flexible form factor placed on the skin to measure arterial diameter changes versus time for a large artery (e.g., the carotid artery). Changes in arterial diameter correspond to the BVP, but are measured from a deeper artery as compared to PPG or tonometry, and thus may be less affected by vasoconstriction. A common measurement site is the carotid artery. Recent work has demonstrated that a blood pressure waveform can be measured from the surface of the skin based on

this principle using a nano-engineered flexible ultrasound array [88]. The device acquires time-varying changes in blood vessel diameter, which are then mapped to an estimate of the underlying blood pressure waveforms. By employing ultrasound to measure this pulsating blood vessel diameter, the device can focus on larger arteries, namely the carotid, which are deeper under the skin than PPGor tonometer-based approaches can access. Accordingly, reduced sensitivity to sensor positioning has been demonstrated as compared to tonometry, and accurate extraction of arterial pressure waveforms has been achieved [88]. Note that this approach requires calibration to acquire the absolute blood pressure values (i.e., systolic, diastolic, and mean arterial pressure), but the waveforms measured are likely the closest to the underlying blood pressure waveforms of the three prior modalities discussed here. An additional concern that should be noted with this approach is that the detection of the artery may require manual positioning and/or image annotation in broad use; however, the array of transducers employed on the device may limit such a need for expert annotation.

#### 4.1.4. Cardio-Mechanical Vibrations

While CRM to date has focused on arterial pulse waveforms measured peripherally, there have also been studies employing cardiogenic vibration signals as an index of hypovolemia based on machine learning techniques in both human subjects (LBNP) [89] and animal models [90]. Note that since these measurements to do not directly yield an arterial pulse waveform, they are not depicted in Figure 5 to avoid confusion. Cardiogenic vibration signals include the seismocardiogram (SCG) and ballistocardiogram (BCG), both of which originate from the vibrations of the chest (SCG) or whole body (BCG) in response to the ejection of blood from the heart and movement of blood through the vasculature [91]. SCG and BCG signals can be measured accurately with inexpensive and commercially available sensors [92,93], and have been demonstrated to be reliable even in the presence of movement [94,95]. As with the other sensing modalities described above, soft, conformal patch based sensing of SCG signals is also possible: Liu, et al. describe an epidermal sensing system for providing mechano-acoustic measurements of cardiovascular health, including heart sounds and SCG signals [96]. Machine learning based analyses performed on these waveforms demonstrated that high quality estimation of blood volume status (analogous to CRM) could be obtained in a pig model of hemorrhage [90]. Importantly, for persons suffering polytraumatic injuries who may not have an available digit or ear, and may have extensive vascular damage that could reduce PPG waveform quality due to increased wave reflections, such cardiogenic vibrations may provide an alternative waveform for CRM-based volume status assessment.

## 4.1.5. Other Emerging Wearable Sensing Devices

The field of wearable sensing has seen a myriad of new devices over the past several years, driven by the use of new materials and fabrication approaches, developments in chemical sensing, and the advent of soft, flexible, and stretchable electronics. These new devices promise to deliver comfortable and high-performance sensing of cardiovascular health parameters with thin, flexible, and stretchable mechanical footprints that resemble the properties of human skin. Emerging technologies of interest also include biodegradable sensors such as the one described in Boutry, et al., for tonometry-based pulse signature sensing [97], and combined chemical/electrophysiological hybrid biosensing systems such as the one presented in Imani, et al. [98]. Additionally, while not discussed in detail here, wearable sensing systems measuring impedance plethysmogram waveforms [99], or magnetic inductance based cardiac waveforms [100], may also be employed.

Table 3 provides a comparison of state-of-the-art sensing technologies for arterial pulse waveform analogs, including summarizing the principle of operation, the typical locations on the body where the signals are captured, and the advantages and disadvantages of each method for application to CRM.


