Abstract
Plethysmographic signals were measured remotely (> 1m) using ambient light and a simple consumer level digital camera in movie mode. Heart and respiration rates could be quantified up to several harmonics. Although the green channel featuring the strongest plethysmographic signal, corresponding to an absorption peak by (oxy-) hemoglobin, the red and blue channels also contained plethysmographic information. The results show that ambient light photo-plethysmography may be useful for medical purposes such as characterization of vascular skin lesions (e. g., port wine stains) and remote sensing of vital signs (e. g., heart and respiration rates) for triage or sports purposes.
Introduction
Detection of the cardio-vascular pulse wave traveling through the body is referred to as Plethysmography (‘Plethysmos’ = increase in Greek) and can be done by means such as variations in air pressure, impedance, or strain. Photo-plethysmography (PPG), introduced in the 1930’s [1] uses light reflectance or transmission and is the least expensive method and simple to use. PPG is based on the principle that blood absorbs light more than surrounding tissue so variations in blood volume affect transmission or reflectance correspondingly.
Applications of PPG include monitoring of oxygen saturation (pulse oxymetry), heart (HR) and respiration (RR) rates, blood pressure, cardiac output, assessment of autonomic functions and detection of peripheral vascular diseases. Remote, non-contact pulse oxymetry and PPG imaging have been explored only relatively recently [2, 3].
To our knowledge, PPG has always been performed with dedicated light sources and typically red and/or infra-red (IR) wavelengths. Due to the historical emphasis of PPG on pulse oxymetry and the associated need to sample relatively deep (e.g., 1 mm) veins and arteries, the visible spectrum (with a shallower penetration depth in skin) has often been ignored as a light source for PPG. Publications describing non-red visible light sources for PPG (e.g. green) are either recent [4, 5] or relatively old [6, 7], and, in all cases, contact probes were used. Ambient visible light is often considered a source of noise [8-11] when using IR light sources and detectors sensitive for IR and visible light.
In this communication, we show that PPG signals can be remotely (several m) measured on the human face with normal ambient light as the source and a simple digital, consumer level photo camera in movie mode. At distances of 1.5 m, signal to noise ratio (SNR) was such that up to four harmonics of the fundamental HR frequency can be measured, thus rendering not only the HR but also the shape of the waveform. In addition to HR and RR monitoring, we show that PPG imaging can be used to characterize regions of high and low pulsatility on facial port wine stains (PWS). PWS are cutaneous capillary malformations characterized by a higher number and/or a larger size of the pathologic capillaries [12]. A better understanding of the hemodynamics involved in PWS is of great interest to improve laser therapeutic outcome.
Our PPG results suggest that not only pulsatility but also phase information regarding the cardio-vascular waveform may be deduced. Although we will attempt to interpret clinical results and discuss their relevance, the main purpose of this paper is to demonstrate remote PPG (imaging) with ambient light. To the best of our knowledge, remote PPG (imaging) with ambient light has never been demonstrated.
Discussion
General
We have shown that movies of the human face, recorded with a simple digital camera and ambient light as the only source of illumination, contain robust reflectance PPG signals. Typically the G channel contains the strongest plethysmographic signal, consistent with the fact that (oxy-) hemoglobin absorbs green light better than red [15] and penetrates sufficiently deeper into the skin as compared to blue light to probe the vasculature. However, it should be noted that the R and B channels may contain complementary information. RR, HR, and their mutual modulation, can be determined. Often, harmonics of the RR and HR are above the noise level which provides information regarding the shape of the plethysmograms. Lateral variations in pulsatility and phase of the cardio-vascular wave have been demonstrated in PWS. With the expressed reservation, we believe that the carotid artery may have been imaged in one volunteer.
Two main issues limit the spatial resolution of the PPG images (power and phase) and accuracy of our current approach. First, movement artifacts and, second, reduced SNR due to CCD generated noise in the recorded pixel values. The first issue may be solvable by improved positioning of the volunteers (often the subjects were standing or sitting without head support), software to laterally synchronize frames and more homogeneous (artificial) illumination to reduce shading artifacts. The second issue may be abetted by simply using more advanced movie cameras with higher pixel resolution, allowing for a good SNR through pixel averaging while maintaining a good spatial resolution in the PPG images.
Artifacts
Although concerns regarding artifacts have been addressed in the results section, some require additional discussion. The movie of PVAC (t) (central panel in media 1) shows that the respiration signal of the cheek is out of phase with the rest of the face (e.g., compare frames for t = 7 and 10s). It is possible that the cheek area is slightly displaced during respiration, and, thus, modulating illumination. The overall RR signal, however, is most likely not artifactual. The RR signal for the three channels have different shapes which is inconsistent with illumination/displacement artifacts. Moreover, the clear AC modulation of RR and HR also indicates that the overall RR signal is a true PPG signal. Shading has been acknowledged as a problem in PPG imaging and is being tackled by advanced technology. Obviously, our PPG imaging approach has yet to date not advanced to these dev elopmental stages.
A similar displacement artifact (now HR related) may underlie at least part of the strong signals in the right carotid artery area. The fact that all three channels (R,G
and B) feature a high power is consistent with this hypothesis. On the other hand, they do show some differential features. In particular, the map for the R channel, (a
wavelength band closer to the wavelengths used for PPG imaging [2]) appears to indicate a vessel like structure, exactly where the carotid artery is embedded. Even if some of the features are displacement artifacts it should be noted that they are basically true plethysmographic signals (measuring volume changes). In fact, tissue pulsatility measured with ultrasound [19] is plethysmography as well. However, the observed features may be not pure PPG as it is commonly understood.
A last source of potential artifacts is that the camera models A570 and A640 have automatic functions to adjust gain during movie recording. Such automatic adjustments
decrease the dynamics of the plethysmographic signals (gain goes up when the subject becomes darker and vice versa). Therefore, it is possible that the signal amplitudes are in fact higher than those measured with these cameras. More worrisome, however, is the fact that the auto-gain functions have a time constant (the adjustment is gradual) and that it is not constant for the entire CCD chip but depends on the local light intensity. If, therefore, the time constant for the auto-gain is different for a PWS area (darker) as compared to normal skin, a phase delay could be induced. Concerned that the HR signal phase differences pointed out are related to this, we tested the A640 for such artifactual phase delays. Animated computer graphics were made to simulate a PWS and adjacent normal skin area (comparable color and pulsatility as in real movies of PWS patients) and recorded with the A640. Phase analysis in a similar way as presented for real PWS herein showed no sign whatsoever that phase changes could be artifactually induced. The A560 model has an option to switch off the auto-gain function which, therefore, is our preferred camera until we locate better cameras for our
purpose.
Remote sensing
All herein presented plethysmographic signals were recorded without additional optical lenses other than the built-in 3x zoom. Since there is interest in remote sensing of HR for combat triage [20] or for athletic monitoring purposes, we tried briefly to use a tele-lens (2x) in combination with one of our digital cameras. At a distance of 11 m between the camera and subject, the signal contained more noise than the signals but the fundamental HR frequency could still be identified very clearly for relatively short time windows of 10s. Using movie camera’s with stronger optical zoom options and higher pixel definitions, we expect to measure HR (and possibly RR) at greater distances than 11 m. This will be the subject of a future study.
Future directions
The strong signals in visible light are surprising in the sense that PPG plethysmography has been practiced for many decades while we found few studies exploiting this band, despite explicit recommendations by Crowe et al. [7] to use green/yellow light for which hemoglobin has absorption peaks [15] (consistent with our finding that the G channel has the strongest plethysmographic signals). This is perhaps due to the strong relationship between plethysmography and pulse oximetry. The latter has traditionally used red and IR wavelengths, one of which is at an isobestic point for oxy- and de-oxyhemoglobin. Although for pulse oximetry these wavelengths are probably the right choice, the demonstrated method has distinct opportunities in other applications. As Gailite [4], Jonsson [21] and Lindberg [6] pointed out, reduction of the sampling depth using shorter wavelengths has advantages if superficial microvasculature rather than deeper vasculature is the focus of PPG. In our research to understand PWS and its response to laser therapy, this is indeed the case. We plan to use spectrally resolved reflectance models [16, 17, 22] and hemodynamic models [23, 24] to understand the measured PPG signals’ relationship to PWS laser therapeutic outcome. Obviously, a more systematic categorization of PPG in PWS compared to adjacent skin as well as relative phase differences needs to be done. Currently, while improving the experimental conditions, we are in the process of designing such a study. Potentially this will reveal useful information, complementary to existing, but relatively complex methods to characterize hemodynamics of PWS such as laser speckle imaging [25], laser Doppler
imaging [26] and Doppler optical coherence tomography [27]. In particular the relative phase differences (Fig. 6) and the differential plethysmogram shapes for PWS and normal skin are intriguing and show promise to help understand PWS hemodynamics beyond their current level.
All of the results in this manuscript concern measurements on the human face. Although we have measured clear PPG signals on other anatomical locations (e.g., wrist, leg) the facial signals were typically stronger, the forehead in particular. For the application of remote detection of vital signals for triage and sports this is helpful because the face is usually uncovered. A robust comparison of PPG signal strengths as a function of anatomical location is a subject for future investigation.
Acknowledgment
This work was made possible through the following NIH funds: AR47551 and EB2495