Comparison of scientific CMOS camera and webcam for monitoring cardiac pulse after exercise

In light of its capacity for remote physiological assessment over a wide range of anatomical locations, imaging photoplethysmography has become an attractive research area in biomedical and clinical community. Amongst recent iPPG studies, two separate research directions have been revealed, i.e., scientific camera based imaging PPG (iPPG) and webcam based imaging PPG (wPPG). Little is known about the difference between these two techniques. To address this issue, a dual-channel imaging PPG system (iPPG and wPPG) using ambient light as the illumination source has been introduced in this study. The performance of the two imaging PPG techniques was evaluated through the measurement of cardiac pulse acquired from the face of 10 male subjects before and after 10 min of cycling exercise. A time-frequency representation method was used to visualize the time-dependent behaviour of the heart rate. In comparison to the gold standard contact PPG, both imaging PPG techniques exhibit comparable functional characteristics in the context of cardiac pulse assessment. Moreover, the synchronized ambient light intensity recordings in the present study can provide additional information for appraising the performance of the imaging PPG systems. This feasibility study thereby leads to a new route for non-contact monitoring of vital signs, with clear applications in triage and homecare.


INTRODUCTION
Photoplethysmography (PPG), as first described in the 1930s 1 , is a simple, low-cost, and non-invasive optical biomonitoring technique used to measure the blood volume changes that occur in the human body due to the pulsatile nature of the circulatory system 2 . Since its modern conception decades ago, its ability to monitor vital physiological parameters in real-time, e.g., blood oxygen saturation, heart and respiration rates, cardiac output and blood pressure, has gained significant attention in the biomedical and clinical community and made PPG a standard of monitoring in various applications. However, the contact sensing modus constrains its practicability in situations of skin healing evaluation, or when free movement is required. One potential way to overcome these problems is to use the recently introduced technique of imaging PPG, which is a remote, contactless diagnostic technique that can assess peripheral blood perfusion [3][4][5][6][7][8] . With significant achievements and improvements of imaging techniques, the past decade has witnessed a prompt growth and substantial accumulation of literature regarding imaging PPG. For instance, Wieringa and colleagues have introduced a multiple wavelength imaging PPG device that provides a possible route toward contactless blood oxygen saturation assessment 4 , and Poh and colleagues have reported a webcam based remote PPG signal acquisition technique using ambient light illumination 6 . These two specific examples indicate two research directions within imaging PPG: scientific camera based imaging PPG (iPPG) and webcam based imaging PPG (wPPG). The former imaging PPG setup, which usually comprises high sensitivity and sample rate and is mostly used with an artificial illumination source, has been proven to be superior in the assessment of multiple physiological parameters, and the latter imaging PPG setup, which normally uses ambient light as the illumination source, shows its advantage in terms of simplicity and low cost. Though successful in acquiring physiological parameters, e.g., heart rate, a number of key questions still remain for the wPPG techni to the accum appraise thei introduced, a shown. Ther light intensity and sensitivi after exercise appraise the i

Subject
A total o from Loughb none were di World Medic to the subjec consuming c study.

Experim
The deta required to si from the righ and ambient Afterwards, Nantik, Mass For each sub camera was probe, webc 640×512 pix different exe i.e., 15 km/h PowerTrek, U were made by

Image p
In previo for extraction were recorde averaging ap recordings w for the CMO set of reduce the pixel valu pixel position filter. Cut-of were then ob

Signal a
Fourier T ability to eas still lead to a generally tra time and freq dependent he asked to perf has been em review, see 9 (SPWVD) w where x(s) and x * (s) are the signal and its complex conjugate respectively, g(s) and h(τ) are two smoothing windows whose effective lengths independently determine the time and frequency smoothing spread. In the present study, Kaiser-Bessel function is adopted as both the time and frequency smoothing window.

Statistical analysis
Significant difference between the physiological measurements under differing conditions was tested with analysis of variance (ANOVA, FACTOR = condition) to demonstrate the influence of exercise on the cardiovascular system. The group difference was evaluated via Posthoc analysis using Duncan's test. Moreover, to test the performance of the two imaging PPG systems, ANOVA (FACTOR = measurement techniques) was performed for comparison between wPPG, iPPG, and cPPG. All analyses were performed with α (Type I error) set at 0.05 using the statistical software program SPSS for Windows, version 17.0.

Physiological measurements
In the present study, a total of four physiological measurements were taken from each of the participants (Fig. 2). Fig.3 summarizes the results of the measured variables: HR, systolic blood pressure (SBP), and diastolic blood pressure (DBP). To statistically analyse the influence of exercise on the cardiovascular system, separate ANOVAs with Post hoc tests were conducted on the obtained variables. The statistical results are shown in Fig.3. Significant effects of exercise have been identified in HR and SBP (F=6.36, p<0.001& F=10.77, p<0.000). Post hoc tests revealed that the HR and SBP after both exercises were significantly higher than those in the rest condition (ex1 vs. rest, p<0.05, ex2 vs. rest, p<0.01). In addition, a significant difference between different exercise levels was also detected in HR and SBP, while the hemodynamic parameters all returned to the rest level after 10 minutes rest. Fig. 4 demonstrates an example of the PPG signals obtained from (a) CMOS camera and (d) web-cam of a single subject (#1) in the rest condition. Additionally, three images were presented to illustrate the decomposed Red, Green, and Blue channels for both cameras ((b) & (e)). Separate PPG signals obtained from each of these three channels were also exhibited, e.g., plethysmographic waveforms are visible in Fig. 4(c) for CMOS camera and (f) for web-cam. Oscillations for HR are most pronounced in the Green channel for the web-cam while the scientific CMOS camera shows comparable PPG signals for all three channels. To estimate the cardiac pulse, the TFR analysis has been employed to reveal the obtained plethysmographic waveform for different techniques. The TFR traces derived from the contact PPG sensor, CMOS camera (Green channel) and webcam (Green channel) have been shown in Fig. 5. It can be seen that the HR frequency (~1.1Hz) and 2 nd harmonic components show close agreement for three measurements.  Fig.6 depicts an example of the 3 min recordings from the wPPG signals together with the light intensity for (a) Green and (b) Red channels. Intensity dependent raw PPG signals have been revealed for both channels. Interestingly, though the light intensity of the Red channel is significantly higher than the Green channel, a pronounced plethysmographic waveform is identified in the Green channel.

Imaging PPG vs. contact PPG
To statistically evaluate the performance of the two imaging techniques, an ANOVA analysis (FACTOR = measurement technique) was employed to compare the difference between these two measurements and the goldstandard contact measurements. The HR wPPG and HR iPPG were obtained through averaging the HR within the TFR GREEN traces, which were calculated from the synchronized 30 sec recordings. In the present study, only the comparison of the rest conditions is considered. Specifically, three physiological monitoring techniques revealed non-significant difference in cardiac pulse detection under rest condition (F=0.01, p=0.9879).

DISCUSSION
A dual-channel imaging PPG system using ambient light illumination has been introduced and evaluated in the present study. The performance of both imaging PPG techniques was appraised by comparing them to a commercial pulse oximetry sensor. Statistical analysis shows no significant difference between these three methods for PPG signal capture, suggesting both imaging PPG systems can successfully obtain information about the cardiovascular variables.
The optimal amount of exercise to maintain fitness and reduce mortality from cardiovascular disease remains a matter of debate 11,12 . Hence, it is a worthwhile endeavour to develop a convenient, remote, and reliable technique to monitor the cardiovascular situation before, during and after exercise. From the physiological measurements obtained via the digital blood pressure meter, heart rate and systolic blood pressure patterns have been shown to be dependent on the intensity of exercise, which agrees well with previous studies 7,8,13,14 . Interestingly, compared to the medium exercise duration (5-min) 7,8 , a longer exercise duration (10-min) in the present study does not trigger a higher response.
It can be seen in Fig. 4 that a pronounced Green channel plethysmographic signal is uncovered, where a similar pattern has been revealed for all subjects. In fact, for most conditions, the light intensity of Red channel is significantly higher than Green channel (Fig.6). This finding is consistent with several webcam and consumer level digital camera imaging PPG studies 5,6 . Actually, it has been indicated in the literature that (oxy-) haemoglobin absorbs more green light than red light and penetrates sufficiently deep into the skin as compared to blue light 5 . Such a pattern is absent from the iPPG signals, which might be related to the relatively similar quantum efficiency for Red, Green, and Blue channels in the camera.
Although the subjects were requested to main motionless during recordings, movements of the subject relative to the camera may occur since the sensor has no contact with the skin. Hence, it is important to recognize the limitations of this study. During the recordings, typical involuntary small movements included mild leaning of the body (and hence the head) towards/away from the camera due to deep breathing, especially after the various exercise conditions. Another limitation is the quantification of the influence of ambient light intensity on the obtained imaging PPG signals. Although the raw PPG signals from the webcam present an intensity dependent pattern, the relationship between the intensity and the PPG signals, e.g., the amplitude of the ac components, is still unclear. A further complete study with full set of subjects and conditions is needed to assess the performance of both imaging PPG techniques in detecting the cardiac pulse after various exercise intensities as well as to quantify the influence of light intensity on the imaging PPG signals.

CONCLUSION
In this study, we have introduced and implemented a dual-channel imaging PPG system, which couples a sensitive scientific CMOS camera and a webcam, and uses ambient light illumination. The performance of both imaging PPG techniques exhibit comparable functional characteristics in detecting the cardiac pulse, indicating a promising alternative to conventional contact PPG. Hence, the results of the present study provide further evidence for webcam-based imaging PPG. Given the low cost and convenience, the webcam leads to a new insight into improving access to medical care and homecare. In comparison, imaging PPG techniques using high speed and high sensitivity cameras maintain a superior capability for assessment of multiple physiological parameters.