Open Access
Add to BibSonomyAbstract
We present experimental evidence that the parameters of green light remitted from a human tissue in-vivo strongly depend on skin contact status. In case when the skin is free of any contact, simultaneous recording of imaging photoplethysmogram (iPPG) and electrocardiogram revealed that contactless iPPG fails in correct estimates of the heart rate in almost half of the cases. Meanwhile, the number of successful correlations between ECG and iPPG is significantly increased when the skin is in contact with a glass plate. These observations are in line with the recently proposed model in which pulsatile arteries deform the connective-tissue components of the dermis thus resulting in temporal modulation of the capillary density interacting with slightly penetrating light.
© 2015 Optical Society of America
1. Introduction
It is well known that visible (or near infrared) light either reflected from or transmitted through a part of the body in-vivo obtains a modulation in time at the heartbeat frequency [1]. It is commonly accepted that such a modulation of light power after its interaction with the living tissue is caused by pulsatile changes of the blood volume in arteries [1–3]. This phenomenon constitutes the basis of photoplethysmography (PPG), a simple, low-cost, and noninvasive technique for monitoring and measuring important parameters of the cardiovascular system performance, such as mean heart rate [4,5], heart rate variability [6,7], blood oxygen level [8,9], and blood pressure [10,11]. PPG devices are used in the routine clinical practice for measuring the oxygen saturation and monitoring the heart rate. However, conventional PPG devices measure the parameters of peripheral blood perfusion only in a single point and operate in physical contact with the subject's skin. To overcome these limitations, noncontact and spatially resolved technology referred to as imaging photoplethysmography (iPPG) was proposed in 2005 by Wieringa et al. [12]. A video camera is used as a photo-receiver in this technique, while a subject is typically illuminated by visible and near infrared light. Since noncontact devices are more preferable in many circumstances, a number of research groups recently started development of iPPG sensors [4,8,13–16]. These devices are intended to reveal at least the same set of physiological parameters as the conventional contact PPG sensors do. However, to our knowledge, there are only few very recent studies devoted to comparison of iPPG devices with the “gold standard” of electrocardiographic (ECG) recordings [4,17–19]. In contrast, the most of iPPG-related researches use the signal from a contact PPG sensor as a reference.
Thus, the aim of the present work was to investigate how well the information about the heart rate (including its variability) revealed from iPPG does correlate with that obtained from ECG in an extended cohort of human subjects. We carried out measurements on both hands of subjects and found that the correlation between contactless iPPG and ECG is insignificant (p > 0.05) in 39 of 86 cases. However, after the hand was brought into contact with a glass plate (a technique which was recently proposed in our group [20]), the number of false readings was diminished to three. We also observed that the amplitude of the reflected light modulation was increased in all subjects after contacting the glass. These findings might find explanation in the frames of a recently proposed model of PPG in which emphasis is given to mechanical deformations of the dermis by pulsating arteries [20]. The extent of such deformation depends on the contact status of the skin.
2. Materials and method
2.1 Participants
A total of 43 apparently healthy subjects (32 males and 11 females) were studied. Age of subjects was from 18 to 74 years. Persons with any neurologic, cardiovascular or skin diseases were not invited to participate in this study. Experiments were performed in the Federal State Institution “1477 Navy Clinical Hospital” of the Ministry of Defense of the Russian Federation, Vladivostok, Russia. This study was conducted in accordance with the standards of application of new medical techniques laid down by Order of the Ministry of Health of the Russian Federation No.25 on 16.02.1994. The study plan was matched to the requirement of the Ethics Committee, 1477 Navy Hospital acted under the Order of Head of the Hospital on 1.12.2014. Ethical approval was obtained prior to the study. All subjects gave their written informed consent of participation in the experiment.
2.2 Instrumentation
We used a custom iPPG system to collect the data from the palm area of a subject. The system was described in detail by Teplov et al. [21]. In summary, we used two conventional light-emitting diodes (LED) operating at a wavelength of 525 nm (green light) to illuminate the subject’s palm. The optical power of each LED was 30 mW and their spectral bandwidth was 60 nm. A digital black-and-white CMOS camera (8-bit model GigE uEye UI-5220SE of Imaging Development Systems GmbH) was used to record videos of the illuminated area, which includes the palm and part of the wrist. All videos were recorded at 30 frames per second (fps) with pixel resolution of 752 × 480 and saved frame-by-frame in PNG format on the laptop. The distance between the camera lens and the palm under study was about 1 m. Experiments were carried out in the laboratory without ambient illumination. ECG was recorded simultaneously with video by a digital electrocardiograph (Fukuda CardiMax FX-7102). To implement ECG recordings, disposable Ag/AgCl electrodes were attached to the left and right wrists with the reference electrodes on the leg.
2.3 Experimental protocol
Before video and ECG recordings, each subject was kept in a quiet laboratory room in a seated position for 15 min. During this period, we measured blood pressure and completed a questionnaire concerning his physical status. During measurements, the subject was seated comfortably with a hand on the glass table as shown in Fig. 1(a). Video camera and illuminating LEDs were placed under the table to provide recordings of the palm through the glass. In the chosen position before video recordings, each subject’s hand was weighed with and without an additional load of 2.0 kg. For each subject’s hand we made two sequential video recordings. First, we recorded 24-s video in conditions of no contact of the palm with the glass, which was provided by putting the wrist and fingers on two soft black supports. The second video of the same length (24 s) was recorded with the subject’s hand onto the glass surface with an additional weight of 2.0 kg distributed over the hand.
Fig. 1 Layout of the experiment and typical distribution of the blood pulsation amplitude. (a) Photograph of the experimental set-up. (b) An example of a single frame excerpt from the recorded video in the case of the palm contacted with the glass. (c) BPA map averaged during four cardiac cycles for the subject’s palm. The color scale on the right shows the amplitude of the pixel value modulation in percent.
2.4. Data processing
Both the recorded video and ECG were processed offline by using custom software implemented in the MATLAB platform. At the first stage we calculated spatial distribution of blood pulsations amplitude (BPA) over the palm area. Algorithm of these calculations is an amplification of the set of the recorded frames synchronously with each cardiac cycle. This technique, referred to as the Blood Pulsation Imaging, was described in detail in our previous publications [15,21]. Briefly, we define the beginning and the end of each cardiac cycle from a PPG waveform obtained after averaging the pixel values within arbitrarily chosen ROI (10 × 10 pixels). Then we generate the reference sinusoidal function for every cardiac cycle, which was used for calculation of the correlation matrix and for mapping the amplitude of blood pulsations [21]. The pulsation amplitude was calculated as an AC/DC amplitude of a PPG waveform for each pixel where AC and DC refer to the alternating (at the heartbeat frequency) and average (slowly varying) portions of the detected photo-signals, respectively. A typical example of a BPA map measured when the palm was contacted with the glass is shown in Fig. 1(c). Such a map was obtained after averaging the calculated amplitude over four cardiac cycles. As one can see, BPA is unevenly distributed over the palm. We found that these maps vary substantially not only from one subject to another but also from the right hand to the left.
At the second stage we found an area with the maximal amplitude in the BPA map (“hot” spot), placed the center of the ROI sizing 5 × 5 pixels into this area, and calculated the PPG waveform in the hot spot by spatial averaging the pixel values within the ROI. This waveform was calculated for the whole length of the video recording (24 s), and it was used for further comparison with the respective ECG recording. To estimate the influence of the skin contact, we first calculated the mean BPA in the hottest spot when the palm was in the contact with the glass, and then compared it with the BPA calculated in the same area of the palm but from the video recorded without any contact. The amplitude of blood pulsations was taken as the peak-to-peak values of the PPG waveform in every cardiac cycle while the individual beat times were estimated from the ECG using positions of the R-peaks. The mean BPA was calculated for all cardiac cycles during 24 s.
We used a custom R-wave detection algorithm to extract the individual beat times from the ECG data. Similarly to the calculated PPG waveform, a peak finding algorithm was used to detect the point corresponding to the beginning of the systole. From these data, we built two time sequences for comparison of ECG with PPG. The first one is the duration of each cardiac cycle as calculated from the positions of R-peaks. The second one is the time delay of the systole position (as estimated from the PPG waveform) in respect to the nearest preceding R-peak normalized with the duration of the respective cardiac cycle. It should be noted that our electrocardiograph has no option of external synchronization. Video and ECG recordings were synchronized manually by the operator who pushed on the start button in both devices by his two fingers. Uncertainty of this synchronization was less than 0.2 s. The definition of the second time series via the delay allowed us to exclude influence of the synchronization uncertainty. Correlation between abovementioned time series was further estimated by using Pearson’s coefficients. The correlation between ECG and PPG was considered as significant with p < 0.005.
3. Results
3.1 Increase of the observed light modulation
In all studied subjects, the amplitude of light modulation at the heartbeat frequency was increased for several folds after the skin was brought into contact with the glass plate. This increase is in full agreement with the PPG model presented in our present work [20]. Typical change of the BPA maps for one of the subjects is presented in Fig. 2. We calculated the BPA map shown in Fig. 2(a) from the video taken when the palm has no contact with the glass plate. In contrast, the map in Fig. 2(c) was obtained from the video of the same subject but with the palm on the glass under the load of 2 kg. It is seen that the amplitude of the pixel-value modulation in the places where the skin contacts the glass is much larger than in other places. The graphs in Fig. 2(b) and 2(d) represent evolution in time of the PPG signals which were measured for free and contacted skin, respectively. These waveforms were calculated by spatial averaging the pixel values within the ROI of 5 × 5 pixels, which corresponds to an area of 1.6 × 1.6 mm2 at the palm. Position of the ROI was chosen in the area of the hot spot in the BPA map for skin contacted with the glass as shown by the black square in Fig. 2(c). In the case of the palm without contact (Fig. 2(a)), the ROI was chosen approximately in the same area of the palm. Red curves in Fig. 2(b) and 2(d) show respective ECG signals.
Fig. 2 BPA maps and PPG waveforms in the “hot” spots. (a) The skin is free of any contact, and (b) the PPG waveform (blue curve) measured in the ROI of 5 × 5 pixels concurrently with the ECG (red curve). (c) The palm is in contact with the glass plate, and (d) the PPG waveform (blue) measured in the ROI placed approximately in the same area of the palm as in (a) shown together with the respective ECG (red). Note that the full scale of PPG axis in graph (d) is 10x that of graph (b). ECG amplitude was kept at the same gain for comparison.
One can clearly see that positions of the maxima in the PPG waveform measured in the case of skin in contact with the glass (Fig. 2(d)) is well correlated (r = 0.95) with positions of the R-wave in the ECG signal. In contrast, such a correlation is less evident (r = 0.44) for noncontact palm (Fig. 2(b)). Moreover, the shape of the PPG waveform in Fig. 2(d) better corresponds to the arterial blood volume modulation with well-resolved reflected wave of the pressure than the waveform in Fig. 2(b).
For data of the particular subject shown in Fig. 2, the amplitude of the light modulation grew up to 7.5 folds after the skin was in contact with the glass plate. Both position of the “hot” spots and BPA in these spots vary substantially from one subject to another. These variations relate to individual features of the palm’s shape that results in uneven distribution of the external pressure affecting the skin in different areas of the contact. In our experiments we had no opportunity to measure the pressure in each individual place of the contact. Nevertheless, we estimated the mean external pressure for each subject by dividing the preliminary measured force (hand’s weight including additional load) on the contact area which was obtained analyzing one of the recorded frames (such as in Fig. 1(b)). The mean contact pressure varied from 27 to 55 mmHg with an average of 35 mmHg calculated over the whole cohort of subjects. Figure 3(a) shows the ratio of the pulsation amplitude in the hot spots measured when the palm contacted the glass to that without contact. The minimal increase of the BPA is 1.5x, while the maximal is 8.4x. In average, we observed fourfold increase of the pulsations amplitude after the skin was in contact with the glass. No correlation (p = 0.8) was found between the rate of the BPA increase upon contact and the age of subjects. In contrast, the amplitude of blood pulsations in the hottest spots (when the palm was free of contact) seems have a tendency to grow up with the subject’s age (p = 0.19, r = 0.14) as it is seen in Fig. 3(b).
Fig. 3 Pulsations in the hot spots as a function of the subject’s age. (a) The ratio of the pulsation amplitude measured with and without contact of the palm with the glass. (b) Blood pulsations amplitude in the “hot” spot when the palm is free of any contact. Solid lines show the tendencies of the graphs.
3.2 Correlation between ECG and PPG in hot spots
Figure 4(a) shows one more example of simultaneous ECG-PPG recordings measured in the hot spot. From the PPG signal (such as shown in Figs. 2(b), 2(d) or 4(a)), we determined the moments at which the waveform reaches the maximal values, which correspond to the systole phase. These moments are marked by red circles in the graph of Fig. 2(d) and by blue circles in Fig. 4. From the ECG recordings (red curve in Fig. 4(a)), we found positions of R peaks in the time scale. These are marked by blue squares in Fig. 4(a) and used for calculation of the cardiac cycle periods. The sequence of the heartbeats periods is shown in Fig. 4(b) by the red curve while the blue curve shows the delay of the systole moments in respect to the R-peaks normalized with the cycle duration. It is seen that both curves have a similar shape and reveal strongly pronounced variability of the heart rate for the particular subject. Statistical calculations show that the Pearson’s coefficient in this case is r = 0.81 with a significance value of p = 0.0000001.
Fig. 4 Comparison of ECG recordings and PPG waveforms. (a) Oscilloscope traces of the ECG signal (red) and PPG waveform (blue) recorded in the “hottest” spot for skin contacted with the glass. (b) Heart-beat periods calculated from the ECG signal (red curve), and the normalized delay between the R peaks and systole peaks from the PPG waveforms (blue curve) as a function of the beat number.
Similar ECG vs PPG comparison was carried out for all 86 hands studied in our experiments. We found out that when the palm was free of contact, the correlation between ECG and PPG was significant (p < 0.05) in only 47 of 86 cases. However, when the skin displacement was limited by the contact with the glass, PPG-waveform sequence has significant correlation with the ECG in 83 of 86 cases. Without contact, the main reason of unsuccessful correlation was mainly related to appearance of spurious peaks in PPG waveforms due to increased influence of the noise at relatively small amplitude. In contrast, for skin contacted with glass even in three exceptional cases of insignificant correlation (p > 0.05) we observed a classical low-noise shape of the PPG waveform. However, the sequence of systole peaks did not follow that of R peaks. The reason for such decorrelation is not clear at the moment.
4. Discussion
The intent of the present study was to demonstrate an important role of mechanical deformations of the dermis in the interaction of light with living biological tissue. This effect becomes apparent in photoplethysmography (PPG), an optical technique which analyses temporal modulation of the light power caused by dynamic processes in the cardiovascular system. Commonly accepted physiological model assumes that the PPG waveform originates from the pulsatile variations of the optical density due to blood-volume pulsations in the arteries [1,2,9]. Blood volume change in capillaries and veins during the time of one cardiac cycle (about 1 s) is considered insignificant [22]. Presented here observations of significant change of the PPG waveform under moderate external pressure on the skin can hardly be explained in the frames of the existing physiological model. Despite of the well-established fact of necessity for skin contact in conventional (one-point) PPG probes, only few investigations of this issue are available in the literature [23–25]. Teng and Zhang investigated the influence of the contacting force on the PPG waveform in the finger and found that the AC/DC ratio of the PPG signal is decreasing with the increasing force [24], which was attributed with change of blood vessels cross-section under external force that was relatively high in their experiments [25]. Only mechanical properties of big vessels, not the properties of other tissue (such as dermis) were considered in the previous model [25]. Researchers working in the field of optical coherence tomography also studied the influence of mechanical contact on light interaction with the tissue [26–28]. They found an increase of the light penetration depth when the tissue is mechanically compressed (the effect referred to as mechanical optical clearing). However, such an increase was observed in conditions of substantial external pressure (> 750 mmHg) and strong pressure gradients [27,28]. Moreover, the clearing effect occurs with a delay of tens of seconds after the compression starts. In contrast, in our experiments the external pressure was significantly smaller (35 mmHg) with negligible pressure gradients. Therefore, neither change of vessels cross-section nor optical clearing could be the reason of the observed increase of the light-modulation amplitude.
Alternatively, significant change of the PPG waveform under low external pressure to the skin can find reasonable explanation within a new physiological model recently proposed by our group [20]. According to this model, pulse oscillations of the arterial transmural pressure, which occur during every cardiac cycle, deform the connective-tissue components of the dermis. During the systole phase, growing transmural pressure of arteries compresses the connective tissue of the dermis in local places depending on the particular anatomy of a subject. Dermis contains both blood and lymphatic capillaries but they are incompressible and do not pulsate at the heartbeat rate [29]. However, due to dermis compression, the distance between adjacent capillaries can be readily changed, which results in modulation of their density synchronously with the arterial pressure in the local place of measurements. Consequently, both the absorption and scattering coefficients from the considered tissue volume are varying during the cardiac cycle, thus leading to temporal modulation of the light intensity returned to the photo-receiver. It means that arterial pulsations are indirectly monitored even by using light which slightly penetrates into the biological tissue, as occurs with the green light at the wavelength of 525 nm used in our experiment [20].
The degree of the dermis compression depends not only on the topology of the arteries, but also on the status of the skin. If the skin has higher elasticity, as it takes place more often among younger subjects, the same transmural pressure in arteries will lead to smaller compression of the dermis of younger compared to elder subjects. In this manner we explain the observed slight increase of the PPG-waveform amplitude with the subject’s age presented in Fig. 3(b). Moreover, local defects of the dermis, such as wounds or tumors could increase the rate of tissue compression by the elastic pulsatile wave thus resulting in a higher amplitude of the light modulation. Similarly, the grade of the dermis compression increases after the skin is led to contact with the glass plate. It results in higher amplitude of stresses in the dermis and consequently in the bigger amplitude of the PPG signal as one can see comparing Fig. 2(b) and Fig. 2(d).
Our model is supported by in vitro measurements of optical properties of skin samples under compression. Chan et al. [23] shown that both the absorption and reduced scattering coefficients of skin samples at the wavelength of 500 nm were increased several times after being compressed by the pressure of 7.5 mmHg. Moreover, such a pressure diminishes the thickness of the samples by two – three folds while it increases the light transmittance through them by less than 10% [23]. On the one hand, this significant change of the tissue thickness under relatively small external pressure could certainly be a reason of capillary density modulation by pulsatile arteries. On the other hand, insignificant increase of the light transmission through the tissue shows that the mechanical optical clearing has minor effect on the light interaction with tissue at small external pressure. In addition we note that recently Fallow et al. [30] reported that the contact pressure in the PPG probe operating with green light was kept at a low level (10 mmHg) to achieve the highest signal-to-noise ratio. However, no reason for this choice was given in this paper [30].
It should be noted that other physiological processes (for example, local muscle constrictions) might also affect the density of the capillary bed. These fluctuations distort the shape of the PPG waveform, produce spurious extremes (such as shown in Fig. 2(b)), and finally impair the correlation between ECG and PPG because of incorrect determination of the intervals for cardiac cycles from the PPG waveform. Capillary-density fluctuations constitute a physiological noise for the PPG signal, which in many cases can be much higher than motion (or other technological) artifacts. We hypothesize that the physiological noise is the main reason of insignificant ECG-vs-PPG correlation for 39 of 86 hands in the case of non-contact video recording. External contacting force increases the rate of transformation of the blood pressure into dermis compression thus increasing SNR of the optical signal. This would explain the observed significant increase of successful ECG-PPG correlations up to 83 of 86 with the hand palm contacting the glass, compared to only 39 of 86 when not contacting the glass.
5. Conclusion and future work
In this paper, we show that the status of human skin affects the light interaction with the dermis in vivo. By contacting the skin with the glass plate under small external pressure (which put a limitation for skin displacement), we modify the temporal parameters of the light remitted from the dermis. At these conditions we observed an increase of the modulation amplitude of blood pulsations on both hands from all 43 studied subjects. In case when the skin is free of any contact, a beat-to-beat sequence derived from iPPG recordings does not correlate with that obtained from ECG in 39 of 86 studied hands. The number of errors in pulse rate estimates by the iPPG system was significantly reduced after the skin was contacted with the glass. On the one hand, our findings show that there are physiological reasons which lead to these errors, rather than motion artifacts which are commonly considered as the major obstacle limiting wide application of PPG devices. On the other hand, our reported observations support a recently proposed model of the PPG-waveform formation [20], which emphasizes the role of elastic deformations of the dermis caused by periodical changes of the arterial transmural pressure during each cardiac cycle. This new model has to be further developed to elaborate the ways of recovering the correct cardiovascular information from fully contactless optical measurements.
Acknowledgments
Financial support by the Russian Science Foundation (grant 15-15-20012) is acknowledged. The authors are grateful to PhD Natalia Podolian and MSc Timofei Efimov for their help in carrying out the experiments.
References and links
1. J. Allen, “Photoplethysmography and its application in clinical physiological measurement,” Physiol. Meas. 28(3), R1–R39 (2007). [CrossRef] [PubMed]
2. S. M. Daly and M. J. Leahy, “‘Go with the flow ’: A review of methods and advancements in blood flow imaging,” J. Biophotonics 6(3), 217–255 (2013). [CrossRef] [PubMed]
3. T. Tamura, Y. Maeda, and M. Yoshida, “Wearable photoplethysmographic sensors—past and present,” Electronics (Basel) 3(2), 282–302 (2014). [CrossRef]
4. Y.-P. Yu, P. Raveendran, and C.-L. Lim, “Dynamic heart rate measurements from video sequences,” Biomed. Opt. Express 6(7), 2466–2480 (2015). [CrossRef] [PubMed]
5. S. Xu, L. Sun, and G. K. Rohde, “Robust efficient estimation of heart rate pulse from video,” Biomed. Opt. Express 5(4), 1124–1135 (2014). [CrossRef] [PubMed]
6. R.-C. Peng, X.-L. Zhou, W.-H. Lin, and Y.-T. Zhang, “Extraction of heart rate variability from smartphone photoplethysmograms,” Comput. Math. Methods Med. 2015, 516826 (2015). [CrossRef] [PubMed]
7. A. Schäfer and J. Vagedes, “How accurate is pulse rate variability as an estimate of heart rate variability? A review on studies comparing photoplethysmographic technology with an electrocardiogram?” Int. J. Cardiol. 166(1), 15–29 (2013). [CrossRef] [PubMed]
8. U. Bal, “Non-contact estimation of heart rate and oxygen saturation using ambient light,” Biomed. Opt. Express 6(1), 86–97 (2015). [CrossRef] [PubMed]
9. T. Aoyagi, “Pulse oximetry: its invention, theory, and future,” J. Anesth. 17(4), 259–266 (2003). [CrossRef] [PubMed]
10. M. Nitzan, Y. Adar, E. Hoffman, E. Shalom, S. Engelberg, I. Z. Ben-Dov, and M. Bursztyn, “Comparison of systolic blood pressure values obtained by photoplethysmography and by Korotkoff sounds,” Sensors (Basel) 13(11), 14797–14812 (2013). [CrossRef] [PubMed]
11. L. Peter, N. Noury, and M. Cerny, “A review of methods for non-invasive and continuous blood pressure monitoring: Pulse transit time method is promising?” IRBM 35(5), 271–282 (2014). [CrossRef]
12. F. P. Wieringa, F. Mastik, and A. F. W. van der Steen, “Contactless multiple wavelength photoplethysmographic imaging: a first step toward “SpO2 camera” technology,” Ann. Biomed. Eng. 33(8), 1034–1041 (2005). [CrossRef] [PubMed]
13. W. Verkruysse, L. O. Svaasand, and J. S. Nelson, “Remote plethysmographic imaging using ambient light,” Opt. Express 16(26), 21434–21445 (2008). [CrossRef] [PubMed]
14. M.-Z. Poh, D. J. McDuff, and R. W. Picard, “Advancements in noncontact, multiparameter physiological measurements using a webcam,” IEEE Trans. Biomed. Eng. 58(1), 7–11 (2011). [CrossRef] [PubMed]
15. A. A. Kamshilin, S. Miridonov, V. Teplov, R. Saarenheimo, and E. Nippolainen, “Photoplethysmographic imaging of high spatial resolution,” Biomed. Opt. Express 2(4), 996–1006 (2011). [CrossRef] [PubMed]
16. M. Kumar, A. Veeraraghavan, and A. Sabharwal, “DistancePPG: Robust non-contact vital signs monitoring using a camera,” Biomed. Opt. Express 6(5), 1565–1588 (2015). [CrossRef] [PubMed]
17. E. B. Blackford and J. R. Estepp, “Effects of frame rate and image resolution on pulse rate measured using multiple camera imaging photoplethysmography,” Proc. SPIE 9417, 94172D (2015). [CrossRef]
18. J. Moreno, J. Ramos-Castro, J. Movellan, E. Parrado, G. Rodas, and L. Capdevila, “Facial video-based photoplethysmography to detect HRV at rest,” Int. J. Sports Med. 36(6), 474–480 (2015). [CrossRef] [PubMed]
19. F. Andreotti, A. Trumpp, H. Malberg, and S. Zaunseder, “Improved heart rate detection for camera-based photoplethysmography by means of Kalman filtering,” in Proceedings of 35th International Conference on Electronics and Nanotechnology (IEEE, 2015), pp. 428–433. [CrossRef]
20. A. A. Kamshilin, E. Nippolainen, I. S. Sidorov, P. V. Vasilev, N. P. Erofeev, N. P. Podolian, and R. V. Romashko, “A new look at the essence of the imaging photoplethysmography,” Sci. Rep. 5(5), 10494 (2015). [CrossRef] [PubMed]
21. V. Teplov, E. Nippolainen, A. A. Makarenko, R. Giniatullin, and A. A. Kamshilin, “Ambiguity of mapping the relative phase of blood pulsations,” Biomed. Opt. Express 5(9), 3123–3139 (2014). [CrossRef] [PubMed]
22. Y. C. Fung, B. W. Zweifach, and M. Intaglietta, “Elastic environment of the capillary bed,” Circ. Res. 19(2), 441–461 (1966). [CrossRef] [PubMed]
23. E. K. Chan, B. Sorg, D. Protsenko, M. O’Neil, M. Motamedi, and A. J. Welch, “Effects of compression on soft tissue optical properties,” IEEE J. Sel. Top. Quantum Electron. 2(4), 943–950 (1996). [CrossRef]
24. X.-F. Teng and Y.-T. Zhang, “The effect of contacting force on photoplethysmographic signals,” Physiol. Meas. 25(5), 1323–1335 (2004). [CrossRef] [PubMed]
25. X.-F. Teng and Y.-T. Zhang, “Theoretical study on the effect of sensor contact force on pulse transit time,” IEEE Trans. Biomed. Eng. 54(8), 1490–1498 (2007). [CrossRef] [PubMed]
26. C. G. Rylander, T. E. Milner, S. A. Baranov, and J. S. Nelson, “Mechanical tissue optical clearing devices: Enhancement of light penetration in ex vivo porcine skin and adipose tissue,” Lasers Surg. Med. 40(10), 688–694 (2008). [CrossRef] [PubMed]
27. C. Drew, T. E. Milner, and C. G. Rylander, “Mechanical tissue optical clearing devices: evaluation of enhanced light penetration in skin using optical coherence tomography,” J. Biomed. Opt. 14(6), 064019 (2009). [CrossRef] [PubMed]
28. M. Y. Kirillin, P. D. Agrba, and V. A. Kamensky, “In vivo study of the effect of mechanical compression on formation of OCT images of human skin,” J. Biophotonics 3(12), 752–758 (2010). [CrossRef] [PubMed]
29. A. Reisner, P. A. Shaltis, D. McCombie, and H. H. Asada, “Utility of the photoplethysmogram in circulatory monitoring,” Anesthesiology 108(5), 950–958 (2008). [CrossRef] [PubMed]
30. B. A. Fallow, T. Tarumi, and H. Tanaka, “Influence of skin type and wavelength on light wave reflectance,” J. Clin. Monit. Comput. 27(3), 313–317 (2013). [CrossRef] [PubMed]
