Estimating heart rate variability using facial video photoplethysmography: a pilot validation study

Estimating heart rate variability using facial video photoplethysmography: a pilot validation study | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Estimating heart rate variability using facial video photoplethysmography: a pilot validation study View ORCID Profile Leszek Pstras , View ORCID Profile Tymoteusz Okupnik , View ORCID Profile Beata Ponikowska , View ORCID Profile Bartlomiej Paleczny doi: https://doi.org/10.1101/2025.02.13.25322028 Leszek Pstras 1 Nalecz Institute of Biocybernetics and Biomedical Engineering, Polish Academy of Sciences , Warsaw, Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Leszek Pstras For correspondence: leszek.pstras{at}ibib.waw.pl Tymoteusz Okupnik 2 Department of Physiology and Pathophysiology, Faculty of Medicine, Wroclaw Medical University , Wroclaw, Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Tymoteusz Okupnik Beata Ponikowska 2 Department of Physiology and Pathophysiology, Faculty of Medicine, Wroclaw Medical University , Wroclaw, Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Beata Ponikowska Bartlomiej Paleczny 2 Department of Physiology and Pathophysiology, Faculty of Medicine, Wroclaw Medical University , Wroclaw, Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Bartlomiej Paleczny Abstract Full Text Info/History Metrics Data/Code Preview PDF ABSTRACT Introduction Video photoplethysmography (vPPG) employs a digital camera to detect blood pulsations in the skin vasculature, which can be used to estimate various physiological parameters. In this study, we investigate the accuracy and precision of two heart rate variability (HRV) indices estimated using a smartphone camera and the facial vPPG technology Shen.AI Vitals. Methods The study group included 35 healthy volunteers (17 females) with median age 25 years (range 20–42 years). The subjects were in a sitting position, keeping their heads relatively still. A smartphone mounted on a tripod was used to acquire 1-min video recordings of participants’ faces. In parallel, a 1-lead chest electrocardiogram (ECG) was recorded to obtain reference values of two analysed time-domain HRV indices: SDNN and lnRMSSD. Results For SDNN, the mean absolute error (MAE) was 3.5 ms (11.0% in relative terms) and the root-mean-square error (RMSE) was 4.5 ms (15.7%). For lnRMSSD, the MAE was 0.24 (7.3%) and RMSE was 0.31 (9.9%). Correlations between the vPPG-based and ECG-based HRV values were strong, with the Pearson correlation coefficient of 0.98 for SDNN and 0.88 for lnRMSSD (P < 0.001 in both cases). Conclusions In a young, white population, the tested vPPG technology estimated HRV indices (SDNN and lnRMSSD) with acceptable accuracy in most subjects, with a slight systematic overestimation, especially for low values. The results should be confirmed in a larger study with greater diversity in age and skin tone. INTRODUCTION Video photoplethysmography (vPPG), also known as remote or imaging photoplethysmography, is an optical technique of recording blood pulsations in the skin vasculature, which can be used to estimate vital signs or other physiological parameters [ 1 , 2 ]. Specifically, vPPG enables recording of fluctuations in the intensity of ambient light reflected from the skin caused by changes in the amount of light absorbed by the blood in the superficial vessels due to their cyclic pulsation. These fluctuations are invisible to the naked eye but usually strong enough to be detected by a digital camera. vPPG provides therefore the possibility of recording PPG signals in a contactless manner without the need of special light sources and using a camera as a light sensor. This makes it possible to use cameras built into mobile devices, such as smartphones, tablets, or laptops, to measure or monitor various physiological parameters [ 3 , 4 ]. Similar to standard PPG signals, vPPG signals can be used to estimate the time intervals between successive blood pulsations, which are relatively similar to the time intervals between heartbeats (although not exactly the same [ 5 , 6 ]). These time intervals can be subsequently used to calculate (estimate) heart rate (HR) as well as various indices of heart rate variability (HRV) describing the variability of the time intervals between successive heartbeats [ 7 ]. HRV is a well-known marker of autonomic nervous system activity with various applications [ 8 , 9 ]. A reduced HRV indicates a less adaptive autonomic regulation of the heart (i.e. an impaired ability to respond to various stimuli) and is an independent risk factor for cardiovascular events and mortality [ 10 , 11 ]. Low HRV is also associated with higher risk for developing hypertension [ 12 , 13 ] or cardiovascular disease [ 14 ]. HRV can also be used to predict the risk of various adverse events, for example the risk of myocardial ischemia in patients without known coronary artery disease [ 15 ], the risk of renal function deterioration in patients with chronic kidney disease [ 16 ], or the risk of falls [ 17 ]. Beyond clinical applications, HRV can be used to monitor and manage the training and recovery process in athletes [ 18 – 20 ], particularly in endurance sports, such as long-distance running, swimming, cycling, rowing, or cross-country skiing [ 19 ], but also in other sports [ 20 – 22 ]. In this study, we investigated HRV indices estimated using facial vPPG technology called Shen.AI Vitals developed by MX Labs (Tallinn, Estonia). This technology employs mobile device cameras to acquire 1-min vPPG signals from several regions of the face (by detecting pulsatile changes in the intensity of ambient light reflected from the facial skin) and then analyses these signals to estimate various physiological parameters, including the HRV index -– SDNN, i.e. the standard deviation of the time intervals between successive normal heartbeats (default), or lnRMSSD, i.e. the natural logarithm of the root mean square of successive differences in time intervals between heartbeats (optional). SDNN is the most common HRV index in the time domain [ 7 , 8 , 23 ] and captures the total variability in inter-beat time intervals in the given time period (quantified by their standard deviation), thus providing an overall measure of current cardiac and autonomic nervous system activity [ 23 , 24 ]. In turn, RMSSD quantifies the variability of successive inter-beat time intervals, thus focusing on high-frequency HR oscillations, reflecting mostly (albeit not exclusively) the activity of the parasympathetic nervous system [ 25 ]. Taking the logarithm of HRV indices is a common method to achieve a more normal distribution [ 7 ]. In particular, the natural logarithm of RMSSD (lnRMSSD) is often used to monitor the vagal-related adaptation to physical training and post-training recovery [ 18 , 19 , 26 ]. The aim of our study was to investigate the accuracy and precision of SDNN and lnRMSSD values estimated by the Shen.AI Vitals algorithms based on vPPG signals recorded with a smartphone camera, as compared with reference values obtained from simultaneously-recorded electrocardiograms (ECG). METHODS Subjects We recruited 38 healthy volunteers (median age 25 years), including 20 females. Written informed consents have been obtained from all participants, and the study was approved by the Bioethics Committee of the Wroclaw Medical University (approval number 227/2022). Data from three subjects had to be excluded due to excessive noise or artefacts in the ECG signals; therefore, the final analysis was based on data from 35 subjects (17 females). For more details on the study participants, see Table 1 . View this table: View inline View popup Download powerpoint Table 1. Characteristics of the 35 study participants. Data presented as median [range]. Study overview Video recordings were taken with the front camera of a smartphone mounted on a tripod at a distance of around 50 cm from participant’s face, at a height adjusted to the given participant (see Fig. 1 ). The subjects were asked to keep their heads steady during the video recording, to breathe normally, and to refrain from speaking. One measurement per participant was performed under resting conditions (two other measurements followed in other conditions but were not analysed in the present study). The study took place in a laboratory room with the window blinds closed and the ceiling lights on. To ensure sufficient light levels, a ring-shaped LED lamp (with a natural light colour) was mounted on the tripod behind the smartphone. The subjects did not engage in any physical activity before participating in the study. During the study, they were sitting with their back supported. Video recordings were initiated around 5 min after connecting all data acquisition devices. Download figure Open in new tab Fig. 1. Measurement setup. The phone and the auxiliary LED lamp were mounted on a tripod at a height adjusted to the given participant sitting approximately 50 cm from the phone. Equipment The front camera of the smartphone (Samsung Galaxy A22) was used for video recordings with the following parameters: video resolution 1080p, f/2.2, frame rate 30 fps. 1-lead chest ECG was recorded simultaneously in Einthoven Lead I configuration (Bio Amp ML132 module and the PowerLab data acquisition system, ADInstruments, Dunedin, New Zealand). Blood pressure was measured oscillometrically using a brachial cuff (Omron M4 Intelli IT). Other measurements (not analysed in the present study) included continuous monitoring of blood pressure (with a finger cuff), breathing rate (using a chest belt), and blood oxygen saturation (with an ear pulse oximeter). Video collection The vPPG mobile app developed by MX Labs (called Heart Monitor) performs a 1-min video recording of the face, which is analysed using the computing power of the mobile device to estimate various physiological parameters displayed to the user during or after the video measurement. The present study was conducted using a special, research mobile app (provided to us by MX Labs), which records a facial video that is subsequently sent to a cloud server to be analysed at a later time using the Shen.AI Vitals algorithms. This was done for two reasons. First, this research app included a special sound signal sent by the app to the PowerLab system at the beginning and end of each video recording to enable synchronization of the ECG signal with the video, which was crucial for our study. Second, using this approach, we were blinded to the values of HRV indices estimated by the tested technology. These values were provided to us by MX Labs only after the study was completed, so that we could compare them with reference values that we had independently calculated from ECG. Since this research app recorded 2-min videos, only the first minute of each recording was analysed in the present study. ECG processing ECG signals were exported to LabChart 8 (ADInstruments) for automatic detection of QRS complexes and determination of the time intervals between successive R peaks (in a few cases, R peaks were marked manually). Data from three subjects had to be excluded due to high levels of noise or artefacts in the ECG signals. The R-R intervals (in ms) and their time stamps were exported to a text file, which was subsequently processed using MATLAB (The Mathworks Inc., USA). First, all R-R intervals contained within the first minute of the video recording were selected. Second, any R-R intervals that were misidentified due to falsely detected R-peaks were corrected by merging two adjacent R-R intervals if their total length was between 80% and 120% of the median length of all identified R-R intervals. To allow for a fair comparison, for the calculation of SDNN, similarly as done in the tested vPPG technology, we removed possible low-frequency trends from the R-R interval time series (R-R tachogram), so that the calculated SDNN value was not affected by possible larger changes in HR during the measurement (note that even though all measurements were taken under resting conditions and the subjects were seated for a few minutes before the measurement, some slow changes in HR during the measurement were still possible and were indeed observed in some of our subjects). To this end, we first interpolated the R-R tachograms to 4 Hz using a piecewise cubic hermite interpolating polynomial (PCHIP) [ 27 ], then we applied the detrending technique proposed by Tarvainen et al. [ 28 ] with lambda = 500, which corresponds to the cut-off frequency of 0.035 Hz [ 29 ], and finally we interpolated the detrended time series back to the original time points (i.e. the end times of the original R-R intervals) to obtain the final detrended tachogram of R-R intervals (in ms, rounded to the whole number). Calculation of reference HRV values SDNN was calculated as the population standard deviation of all R-R intervals in the analysed 1-min period (from the detrended R-R tachogram). Given the manual and/or automatic correction of falsely detected R-peaks, the analysed intervals were treated as corresponding to normal heartbeats. The second HRV index, i.e. lnRMSSD, was calculated as the natural logarithm of the root mean square of successive differences in R-R intervals in the analysed 1-min period (here, any potential low-frequency trends would have none or minimal impact on the results, and hence we analysed the original, non-detrended R-R tachogram in order not to introduce any potential errors in the process of signal interpolation). All reference HRV values were rounded similarly as the tested values, i.e. to the nearest whole number for SDNN (in ms) and to one tenth for lnRMSSD (dimensionless). Statistics We compared the values of HRV indices estimated by the tested vPPG technology with the reference ECG-based values using mean errors (ME), standard deviations of errors (SD), mean absolute errors (MAE), and root-mean-square errors (RMSE) in both absolute terms and relative terms, i.e. with respect to reference values. Correlation between the tested and reference values was assessed using the Pearson correlation coefficient (r). Statistical significance was set at P = 0.05. The agreement between the tested and reference values was visualized using scatter plots. RESULTS Table 2 presents a summary of HR values and the values of the two analysed HRV indices (SDNN and lnRMSSD) determined in the study participants based on ECG recordings. In Table 3 we show various measures of accuracy and precision of the two HRV indices estimated using the tested vPPG technology (compared to ECG-based values), in both absolute and relative terms, i.e. with respect to reference values. View this table: View inline View popup Download powerpoint Table 2. Summary of ECG-based values of average heart rate (HR) and 1-min heart rate variability indices across all study participants (n=35). SDNN – standard deviation of normal heartbeat time intervals, lnRMSSD – natural logarithm of the root mean square of successive differences in time intervals between heartbeats. View this table: View inline View popup Download powerpoint Table 3. Measures of accuracy and precision of two heart rate variability indices estimated using the Shen.AI Vitals technology as compared with ECG-based values, expressed in absolute terms and in relative terms (as percentage of ECG-based values). Abbreviations: SDNN – standard deviation of normal heartbeat time intervals, lnRMSSD – natural logarithm of the root mean square of successive differences in time intervals between heartbeats, ME – mean error, SD – standard deviation of errors, MAE – mean absolute error, RMSE – root-mean-square error, MPE – mean percentage error, SDPE – standard deviation of percentage errors, MAPE – mean absolute percentage error, RMSPE – root-mean-square percentage error. Fig. 2 presents the vPPG-based HRV indices plotted against the values obtained from ECG. For both SDNN and lnRMSSD, the correlation between the values obtained by the two methods was very strong (r = 0.98 for SDNN and r = 0.88 for lnRMSSD, P < 0.001 in both cases). Download figure Open in new tab Fig. 2. Heart rate variability indices estimated by the Shen.AI Vitals technology during 1-min facial video measurements plotted against reference values obtained from simultaneously recorded electrocardiogram (ECG). A) Standard deviation of the time intervals between normal heartbeats (SDNN). B) Natural logarithm of the root mean square of successive differences in time intervals between heartbeats (lnRMSSD). Dotted lines indicate arbitrarily chosen error limits. Abbreviations: MAE – mean absolute error, MAPE – mean absolute percentage error. In almost all cases (34 out of 35), SDNN values estimated by the tested technology were within 10 ms of the ECG-based reference values (the remaining case had an error of 11 ms). For lnRMSSD, almost 90% of vPPG-based values (31 out of 35) were within 0.5 of the reference values. For SDNN, the mean error between the vPPG-based and ECG-based values was 2.8 ms (9.6% in relative terms), suggesting a slight systematic overestimation of SDNN by the tested technology, particularly for low SDNN values. Similarly, we observed overestimation of lnRMSSD values by around 0.2 (7.2% in relative terms), especially for low values. In the vast majority of cases, absolute percentage errors (APE) did not exceed 10% for both SDNN and lnRMSSD (see Table 4 and Fig. 3 ). Mean absolute percentage errors (MAPE) were 11.0% and 7.3% for SDNN and lnRMSSD, respectively. In the case of lnRMSSD, errors never exceeded 25%, whereas in the case of SDNN, this was true for 31 subjects (89%) – see Table 4 . Download figure Open in new tab Fig. 3. Errors in heart rate variability indices estimated by the Shen.AI Vitals technology from 1-min facial video recordings in 35 subjects (as compared with ECG-based values). A) Absolute percentage errors (APE) in SDNN (blue bar, left) and lnRMSSD (red bar, right). B) Absolute percentage errors (APE) in SDNN (blue bar, left) and absolute errors (AE) in SDNN (green bar, right). Abbreviations: SDNN – standard deviation of normal heartbeat time intervals, lnRMSSD – natural logarithm of the root mean square of successive differences in time intervals between heartbeats. View this table: View inline View popup Download powerpoint Table 4. Distribution of mean absolute percentage errors (MAPE) in SDNN and lnRMSSD values estimated by the Shen.AI Vitals technology in 35 studied subjects as compared to reference values obtain from simultaneously recorded electrocardiogram. DISCUSSION In this study, we investigated the accuracy and precision of two popular time-domain HRV indices – SDNN and lnRMSSD – estimated by the tested vPPG technology based on a 1-min facial video taken with a smartphone camera. HRV indices have been traditionally determined based on 5-min measurements. However, shorter measurements, referred to as ultra short-term HRV, are becoming more and more common [ 30 , 31 ]. For RMSSD, it has been shown that even measurements shorter than 60 s should be sufficient to obtain results similar to standard 5-min measurements [ 32 – 34 ], whereas for SDNN the recommendations vary from 30 s [ 32 ], through 60 s [ 35 ], up to 4 min [ 33 ]. Our results The correlation between the vPPG-based and ECG-based SDNN values was very high (r = 0.98) and the mean absolute error (MAE) was relatively low (3.5 ms); however, the mean absolute percentage error (MAPE) was slightly above 10% (11.0%). For lnRMSSD, the correlation with ECG-based values was somewhat lower (r = 0.88), which is mainly due to relatively lower range of the observed values, with MAE of 0.2 and MAPE below 10% (7.3%). For SDNN, in a few subjects we observed relatively high percentage errors, exceeding 40% in two cases. However, in these cases, SDNN was relatively low, and hence in absolute terms these errors did not exceed 10 ms (see Fig. 3 ). For lnRMSSD, the relative accuracy was better, as the errors never exceeded 25%. Previous studies Similar results for SDNN (r = 0.98, MAE = 3.5 ms) were found in a study of 15 subjects, in whom vPPG-based values were compared to values obtained from classic finger PPG [ 36 ]. A slightly higher MAE (4.9 ms) was reported in a study on 14 subjects (with classic PPG as reference) [ 37 ], while a slightly lower MAE (2.7 ms) was reported in another vPPG study on 5 subjects (with chest-band ECG as reference) [ 38 ]. Much higher values of MAE (8.1 ms [ 39 ] and 18.0 ms [ 40 ]) were reported in studies based on a publicly available dataset with 59 video recordings of 10 subjects (with finger PPG as reference), although that dataset includes also cases where participants were asked to talk or move their head during the recordings [ 41 ]. In another study that used several publicly available datasets, MAE for SDNN was reported to range from 6.2 to 17.5 ms [ 42 ]. Note, however, that the results of different vPPG studies should be compared with great caution, not only because of possible differences in the behaviour and stability of the subjects, but also because of probable differences in the characteristics of the studied subjects (in terms of age, skin tone, or possible makeup), as well as because of the possible differences in terms of lighting, recording equipment, or the measurement setup (e.g. body position or distance from the camera). Also, the length of the measurement (video recording) may greatly influence the agreement between vPPG-based and reference HRV values. In particular, the agreement for SDNN is expected to increase with the recording length [ 39 ], because the longer the measurement, the higher the SDNN [ 33 ], and for higher values the errors are typically lower. For instance, a lower MAE for SDNN (2.8 ms) was reported in a study on 55 subjects involving 5-min video recordings with average SDNN of 48 ms (vs 41 ms in our study) [ 43 ]. On the other hand, in another study with 5-min video recordings in 8 subjects, errors in SDNN were higher than in our study (mean error 6.7 ms), despite the similar average level of SDNN (41 ms) [ 44 ]. For lnRMSSD, we have not found any studies reporting the accuracy of vPPG-based measurements; however, there have been several studies investigating the accuracy of lnRMSSD from contact-based PPG measurements, which in theory are expected to be more reliable. For instance, a higher correlation of finger infrared PPG-based lnRMSSD with ECG-based values (r = 0.98) and a lower mean error (around 0.05) was reported in a study on 30 subjects [ 45 ], although in that study the average value of lnRMSSD was higher than in our study (3.9 vs 3.5), which, again, could have affected the level of accuracy. In another study on 31 subjects with finger camera-based PPG measurements, the mean error in lnRMSSD (as compared to ECG-based values) was similar to that in our study (0.2-0.3 in four different sessions), but the correlation coefficient was much lower (r = 0.59-0.76), which could have been due to the fact that the participants were not exclusively white [ 46 ]. In a study on 39 subjects investigating wrist-based PPG measurements of lnRMSSD (vs chest-band ECG), MAE was similar to that in our study (around 0.2), but MAPE was somewhat lower (5.6%) [ 47 ], although they studied 5-min measurements and the lnRMSSD values were, again, higher than in our study (likely due to supine position [ 48 ]). Overestimation of HRV indices For both SDNN and lnRMSSD we observed a positive bias, i.e. overestimation of HRV indices by the tested technology, particularly for low values. These systematic differences between the vPPG-based and ECG-based HRV indices may be explained by the fact that the peaks in the vPPG signals are generally more difficult to identify compared to R peaks in ECG [ 49 ]. This is partly because the peaks in blood pulsations are generally less sharp compared to R peaks and partly because the vPPG signals are typically recorded with much more noise and artefacts compared to ECG and at a much lower frequency, although it has been suggested that the frequency of 25 Hz should be sufficient for the estimation of the most common HRV indices [ 50 ] (in our study, the vPPG signals were obtained at 30 Hz). Any errors in detecting peaks in blood pulsations translate into errors in the estimated time intervals between successive blood pulsations. This inevitably leads to underestimating the length of some interbeat time intervals and overestimating the length of other time intervals, which in consequence leads to overestimation of HRV. The fact that we observed greater overestimation for low HRV values is expected, given that at low HRV the time intervals between successive blood pulsations are generally more similar to each other (compared to the case with high HRV), and hence any error in estimating individual interbeat time intervals introduces a relatively higher error in the given HRV index. Similar overestimation of HRV by vPPG, particularly for low values, has been reported in the literature [ 36 , 44 , 45 , 47 ]. PRV vs HRV It should be noted that the time intervals analysed by vPPG technology are the time intervals between successive blood pulsations in the skin vasculature, which are not exactly the same as the time intervals between the corresponding R peaks in the ECG signal. Therefore, vPPG technology measures in fact not the heart rate variability but the pulse rate variability (PRV). It has been shown in several studies that HRV and PRV indices are not necessarily equivalent [ 5 , 6 ], which can be explained by the fact that PRV depends not only on HRV but also on the variability of stroke volume, blood pressure, and the speed of pulse wave propagation [ 51 ]. It has been even suggested that PRV should be treated separately from HRV [ 52 ]. However, since the tested technology reports PRV as HRV, we compared the tested values to ECG-based HRV values, thus effectively assessing the accuracy of the tested technology in estimating HRV. Besides, it has been shown that in young and healthy people in resting conditions (as investigated in our study), PRV is generally a good substitute of HRV [ 49 , 53 – 57 ]. Limitations There are some limitations of our study that should be noted. First, we studied a relatively small cohort, which included predominantly young subjects. Although we expect no significant differences in terms of accuracy of vPPG-based HR estimation in older subjects, the age factor may be important for the accuracy of estimating HRV indices. It is well known that HRV decreases with age [ 58 – 61 ], and hence, given the relatively higher bias of vPPG technology for low HRV values, the level of accuracy of HRV estimation in older subjects could be accordingly lower. Second, we studied only white subjects, in whom vPPG technology is known to perform better compared to dark-skinned individuals [ 62 , 63 ]. Finally, our study was conducted in a laboratory setting, with controlled lighting, the smartphone mounted on a tripod at participants’ head level, and the participants were required to keep their heads in a stable position and refrain from speaking or changing facial expression. Therefore, our results are not necessarily generalizable to other, more natural settings, with potentially lower facial illumination levels or a less stable head position. CONCLUSIONS In young, white subjects in the sitting position, 1-min SDNN and lnRMSSD values estimated using the tested facial vPPG technology showed acceptable accuracy in most subjects (as compared with ECG-based values), with mean absolute percentage errors of 11.0% and 7.3%, respectively, and a slight systematic overestimation, especially for low HRV values. These results should be confirmed in a larger study with greater diversity in terms of participants’ age and skin tone. Data Availability Anonymized data on the heart rate variability indices analyzed in the present study are available upon reasonable request to the corresponding author. Conflicts of interest The study was commissioned by MX Labs and conducted under a contractual agreement between MX Labs and Wroclaw Medical University. LP and TO report personal fees from MX Labs. Other authors have no conflicts of interest. Author contributions L. Pstras: Conceptualization, Methodology, Software, Data curation, Formal Analysis, Writing – original draft, Writing – review & editing, Visualization. T. Okupnik: Investigation, Data curation. B. Ponikowska: Resources, Supervision. B. Paleczny: Conceptualization, Supervision, Writing – review & editing. Acknowledgements MX Labs has provided us with a special, research version of its mobile application as well as the smartphone and the tripod with an LED lamp used in the study. REFERENCES [1]. ↵ Molinaro N , Schena E , Silvestri S , Bonotti F , Aguzzi D , Viola E , et al. Contactless Vital Signs Monitoring From Videos Recorded With Digital Cameras: An Overview . Frontiers in Physiology . 2022 ; 13 : 801709 . [2]. ↵ van Es VAA , Lopata RGP , Scilingo EP , Nardelli M . Contactless Cardiovascular Assessment by Imaging Photoplethysmography: A Comparison with Wearable Monitoring . Sensors (Basel ). 2023 ; 23 :1505. [3]. ↵ Pham C , Poorzargar K , Nagappa M , Saripella A , Parotto M , Englesakis M , et al. Effectiveness of consumer-grade contactless vital signs monitors: a systematic review and meta-analysis . J Clin Monit Comput . 2022 ; 36 : 41 – 54 . OpenUrl CrossRef PubMed [4]. ↵ Selvaraju V , Spicher N , Wang J , Ganapathy N , Warnecke JM , Leonhardt S , et al. Continuous Monitoring of Vital Signs Using Cameras: A Systematic Review . Sensors . 2022 ; 22 :4097. [5]. ↵ Mejia-Mejia E , Budidha K , Abay TY , May JM , Kyriacou PA . Heart Rate Variability (HRV) and Pulse Rate Variability (PRV) for the Assessment of Autonomic Responses . Front Physiol . 2020 ; 11 : 779 . [6]. ↵ Iozzia L , Cerina L , Mainardi L . Relationships between heart-rate variability and pulse-rate variability obtained from video-PPG signal using ZCA . Physiological Measurement . 2016 ; 37 : 1934 – 44 . OpenUrl PubMed [7]. ↵ Shaffer F , Ginsberg JP . An Overview of Heart Rate Variability Metrics and Norms . Front Public Health . 2017 ; 5 : 258 . [8]. ↵ Heart rate variability: standards of measurement, physiological interpretation and clinical use . Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology . Circulation . 1996 ;93:1043-65. [9]. ↵ Billman GE , Huikuri HV , Sacha J , Trimmel K . An introduction to heart rate variability: methodological considerations and clinical applications . Front Physiol . 2015 ; 6 : 55 . [10]. ↵ Shaffer F , McCraty R , Zerr CL . A healthy heart is not a metronome: an integrative review of the heart’s anatomy and heart rate variability . Frontiers in Psychology . 2014 ; 5 :1040. [11]. ↵ Singh N , Moneghetti KJ , Christle JW , Hadley D , Froelicher V , Plews D. Heart Rate Variability: An Old Metric with New Meaning in the Era of Using mHealth technologies for Health and Exercise Training Guidance. Part Two: Prognosis and Training . Arrhythm Electrophysiol Rev . 2018 ; 7 : 247 – 55 . OpenUrl PubMed [12]. ↵ Singh JP , Larson MG , Tsuji H , Evans JC , O’Donnell CJ , Levy D . Reduced heart rate variability and new-onset hypertension: insights into pathogenesis of hypertension: the Framingham Heart Study . Hypertension . 1998 ; 32 : 293 – 7 . OpenUrl CrossRef [13]. ↵ Schroeder EB , Liao D , Chambless LE , Prineas RJ , Evans GW , Heiss G . Hypertension, blood pressure, and heart rate variability: the Atherosclerosis Risk in Communities (ARIC) study . Hypertension . 2003 ; 42 : 1106 – 11 . OpenUrl CrossRef [14]. ↵ Kubota Y , Chen LY , Whitsel EA , Folsom AR . Heart rate variability and lifetime risk of cardiovascular disease: the Atherosclerosis Risk in Communities Study . Annals of Epidemiology . 2017 ; 27 : 619 – 25 .e2. OpenUrl PubMed [15]. ↵ Goldenberg I , Goldkorn R , Shlomo N , Einhorn M , Levitan J , Kuperstein R , et al. Heart Rate Variability for Risk Assessment of Myocardial Ischemia in Patients Without Known Coronary Artery Disease: The HRV-DETECT (Heart Rate Variability for the Detection of Myocardial Ischemia) Study† . Journal of the American Heart Association . 2019 ; 8 : e014540 . OpenUrl PubMed [16]. ↵ Chou YH , Huang WL , Chang CH , Yang CCH , Kuo TBJ , Lin SL , et al. Heart rate variability as a predictor of rapid renal function deterioration in chronic kidney disease patients . Nephrology (Carlton ). 2019 ; 24 : 806 – 13 . OpenUrl PubMed [17]. ↵ Castaldo R , Melillo P , Izzo R , De Luca N , Pecchia L . Fall Prediction in Hypertensive Patients via Short-Term HRV Analysis . IEEE J Biomed Health Inform . 2017 ; 21 : 399 – 406 . OpenUrl PubMed [18]. ↵ Bellenger CR , Fuller JT , Thomson RL , Davison K , Robertson EY , Buckley JD . Monitoring Athletic Training Status Through Autonomic Heart Rate Regulation: A Systematic Review and Meta-Analysis . Sports Medicine . 2016 ; 46 : 1461 – 86 . OpenUrl CrossRef PubMed [19]. ↵ Plews DJ , Laursen PB , Stanley J , Kilding AE , Buchheit M . Training adaptation and heart rate variability in elite endurance athletes: opening the door to effective monitoring . Sports Med . 2013 ; 43 : 773 – 81 . OpenUrl CrossRef PubMed [20]. ↵ Nakamura FY , Flatt AA , Pereira LA , Ramirez-Campillo R , Loturco I , Esco MR . Ultra-Short-Term Heart Rate Variability is Sensitive to Training Effects in Team Sports Players . J Sports Sci Med . 2015 ; 14 : 602 – 5 . OpenUrl PubMed [21]. Nakamura FY , Antunes P , Nunes C , Costa JA , Esco MR , Travassos B . Heart Rate Variability Changes From Traditional vs. Ultra-Short-Term Recordings in Relation to Preseason Training Load and Performance in Futsal Players . Journal of strength and conditioning research . 2020 ; 34 : 2974 – 81 . OpenUrl PubMed [22]. ↵ Chen Y-S , Pagaduan JC , Bezerra P , Crowley-McHattan ZJ , Kuo C-D , Clemente FM . Agreement of Ultra-Short-Term Heart Rate Variability Recordings During Overseas Training Camps in Under-20 National Futsal Players . Frontiers in Psychology . 2021 ; 12 . [23]. ↵ Rajendra Acharya U , Paul Joseph K , Kannathal N , Lim CM , Suri JS . Heart rate variability: a review . Medical and Biological Engineering and Computing . 2006 ; 44 : 1031 – 51 . OpenUrl [24]. ↵ Thayer JF , Yamamoto SS , Brosschot JF . The relationship of autonomic imbalance, heart rate variability and cardiovascular disease risk factors . Int J Cardiol . 2010 ; 141 : 122 – 31 . OpenUrl CrossRef PubMed Web of Science [25]. ↵ Coumel P , Maison-Blanche P , Catuli D . Heart rate and heart rate variability in normal young adults . J Cardiovasc Electrophysiol . 1994 ; 5 : 899 – 911 . OpenUrl CrossRef PubMed [26]. ↵ Schmitt L , Regnard J , Millet GP . Monitoring Fatigue Status with HRV Measures in Elite Athletes: An Avenue Beyond RMSSD? Frontiers in Physiology . 2015 ; 6 : 343 . [27]. ↵ Fritsch FN , Carlson RE . Monotone Piecewise Cubic Interpolation . SIAM Journal on Numerical Analysis . 1980 ; 17 : 238 – 46 . OpenUrl CrossRef Web of Science [28]. ↵ Tarvainen MP , Ranta-aho PO , Karjalainen PA . An advanced detrending method with application to HRV analysis . IEEE Transactions on Biomedical Engineering . 2002 ; 49 : 172 – 5 . OpenUrl CrossRef PubMed Web of Science [29]. ↵ Tarvainen MP , Niskanen JP , Lipponen JA , Ranta-Aho PO , Karjalainen PA . Kubios HRV-heart rate variability analysis software . Comput Methods Programs Biomed . 2014 ; 113 : 210 – 20 . OpenUrl CrossRef PubMed [30]. ↵ Pecchia L , Castaldo R , Montesinos L , Melillo P . Are ultra-short heart rate variability features good surrogates of short-term ones? State-of-the-art review and recommendations . Healthcare Technology Letters . 2018 ; 5 : 94 – 100 . OpenUrl PubMed [31]. ↵ Shaffer F , Meehan ZM , Zerr CL . A Critical Review of Ultra-Short-Term Heart Rate Variability Norms Research . Front Neurosci . 2020 ; 14 : 594880 . [32]. ↵ Munoz ML , van Roon A , Riese H , Thio C , Oostenbroek E , Westrik I , et al. Validity of (Ultra-)Short Recordings for Heart Rate Variability Measurements . PLoS One . 2015 ; 10 : e0138921 . OpenUrl CrossRef PubMed [33]. ↵ Baek HJ , Cho C-H , Cho J , Woo J-M . Reliability of ultra-short-term analysis as a surrogate of standard 5-min analysis of heart rate variability . Telemed J E Health . 2015 ; 21 : 404 – 14 . OpenUrl CrossRef PubMed [34]. ↵ Esco MR , Flatt AA . Ultra-Short-Term Heart Rate Variability Indexes at Rest and Post-Exercise in Athletes: Evaluating the Agreement with Accepted Recommendations . J Sports Sci Med . 2014 ; 13 : 535 – 41 . OpenUrl PubMed [35]. ↵ Shaffer F , Shearman S , Meehan ZM . The Promise of Ultra-Short-Term (UST) Heart Rate Variability Measurements . Biofeedback . 2016 ; 44 : 229 – 33 . OpenUrl [36]. ↵ Melchor Rodríguez A , Ramos-Castro J . Video pulse rate variability analysis in stationary and motion conditions . BioMedical Engineering OnLine . 2018 ; 17 : 11 . [37]. ↵ Odinaev I , Prae-Arporn K , Wong KL , Chin JW , Chan TT , Goyal R , et al. Camera-based heart rate variability and stress measurement from facial videos. 2022 IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE) 2022 . p. 168-9. [38]. ↵ Huang R-Y , Dung L-R . Measurement of heart rate variability using off-the-shelf smart phones . BioMedical Engineering OnLine . 2016 ; 15 : 11 . [39]. ↵ Finzgar M , Podrzaj P . Feasibility of assessing ultra-short-term pulse rate variability from video recordings . PeerJ . 2020 ; 8 : e8342 . OpenUrl PubMed [40]. ↵ Gudi A , Bittner M , van Gemert J . Real-Time Webcam Heart-Rate and Variability Estimation with Clean Ground Truth for Evaluation . Applied Sciences . 2020 ; 10 :8630. [41]. ↵ Stricker R , Müller S , Gross H . Non-contact video-based pulse rate measurement on a mobile service robot . The 23rd IEEE International Symposium on Robot and Human Interactive Communication 2014 . p. 1056-62. [42]. ↵ Odinaev I , Wong KL , Chin JW , Goyal R , Chan TT , So RHY . Robust Heart Rate Variability Measurement from Facial Videos . Bioengineering . 2023 ; 10 : 851 . [43]. ↵ Sakamoto T , Nakamura T . Noncontact sensing of heart rate variability from facial video using the topology algorithm . IEICE Communications Express . 2022 ; 11 : 313 – 8 . OpenUrl [44]. ↵ Moreno J , Ramos-Castro J , Movellan JR , Parrado E , Rodas G , Capdevila L . Facial Video-Based Photoplethysmography to Detect HRV at Rest . International Journal of Sports Medicine . 2015 ; 36 : 474 – 80 . OpenUrl PubMed [45]. ↵ Esco MR , Flatt AA , Nakamura FY . Agreement Between a Smartphone Pulse Sensor Application and Electrocardiography for Determining lnRMSSD . The Journal of Strength & Conditioning Research . 2017 ; 31 : 380 – 5 . OpenUrl PubMed [46]. ↵ Holmes CJ , Fedewa MV , Winchester LJ , MacDonald HV , Wind SA , Esco MR . Validity of Smartphone Heart Rate Variability Pre- and Post-Resistance Exercise . Sensors . 2020 ; 20 :5738. [47]. ↵ Nuuttila O-P , Korhonen E , Laukkanen J , Kyröläinen H . Validity of the Wrist-Worn Polar Vantage V2 to Measure Heart Rate and Heart Rate Variability at Rest . Sensors . 2021 ; 22 : 137 . [48]. ↵ da Cruz CJG , Porto LGG , da Silva Rolim P , de Souza Pires D , Garcia GL , Molina GE. Impact of heart rate on reproducibility of heart rate variability analysis in the supine and standing positions in healthy men . Clinics (Sao Paulo) . 2019 ; 74 : e806 . OpenUrl [49]. ↵ Schäfer A , Vagedes J . How accurate is pulse rate variability as an estimate of heart rate variability?: A review on studies comparing photoplethysmographic technology with an electrocardiogram . International Journal of Cardiology . 2013 ; 166 : 15 – 29 . OpenUrl CrossRef PubMed [50]. ↵ Choi A , Shin H . Photoplethysmography sampling frequency: pilot assessment of how low can we go to analyze pulse rate variability with reliability? Physiol Meas . 2017 ; 38 : 586 – 600 . OpenUrl PubMed [51]. ↵ Constant I , Laude D , Murat I , Elghozi JL . Pulse rate variability is not a surrogate for heart rate variability . Clin Sci (Lond ). 1999 ; 97 : 391 – 7 . OpenUrl CrossRef PubMed [52]. ↵ Yuda E , Shibata M , Ogata Y , Ueda N , Yambe T , Yoshizawa M , et al. Pulse rate variability: a new biomarker, not a surrogate for heart rate variability . Journal of Physiological Anthropology . 2020 ; 39 : 21 . [53]. ↵ McKinley PS , Shapiro PA , Bagiella E , Myers MM , De Meersman RE , Grant I , et al. Deriving heart period variability from blood pressure waveforms . J Appl Physiol (1985) . 2003 ; 95 : 1431 – 8 . OpenUrl CrossRef PubMed [54]. Khandoker AH , Karmakar CK , Palaniswami M . Comparison of pulse rate variability with heart rate variability during obstructive sleep apnea . Med Eng Phys . 2011 ; 33 : 204 – 9 . OpenUrl PubMed [55]. Lu S , Zhao H , Ju K , Shin K , Lee M , Shelley K , et al. Can photoplethysmography variability serve as an alternative approach to obtain heart rate variability information? J Clin Monit Comput . 2008 ; 22 : 23 – 9 . OpenUrl CrossRef PubMed [56]. Lu G , Yang F , Taylor JA , Stein JF . A comparison of photoplethysmography and ECG recording to analyse heart rate variability in healthy subjects . J Med Eng Technol . 2009 ; 33 : 634 – 41 . OpenUrl CrossRef PubMed [57]. ↵ Selvaraj N , Jaryal A , Santhosh J , Deepak KK , Anand S . Assessment of heart rate variability derived from finger-tip photoplethysmography as compared to electrocardiography . J Med Eng Technol . 2008 ; 32 : 479 – 84 . OpenUrl CrossRef PubMed [58]. ↵ Pagani M , Lombardi F , Guzzetti S , Rimoldi O , Furlan R , Pizzinelli P , et al. Power spectral analysis of heart rate and arterial pressure variabilities as a marker of sympatho-vagal interaction in man and conscious dog . Circ Res . 1986 ; 59 : 178 – 93 . OpenUrl Abstract / FREE Full Text [59]. Bonnemeier H , Richardt G , Potratz J , Wiegand UK , Brandes A , Kluge N , et al. Circadian profile of cardiac autonomic nervous modulation in healthy subjects: differing effects of aging and gender on heart rate variability . J Cardiovasc Electrophysiol . 2003 ; 14 : 791 – 9 . OpenUrl CrossRef PubMed Web of Science [60]. Ramaekers D , Ector H , Aubert AE , Rubens A , Van de Werf F . Heart rate variability and heart rate in healthy volunteers. Is the female autonomic nervous system cardioprotective? Eur Heart J . 1998 ; 19 : 1334 – 41 . OpenUrl CrossRef PubMed Web of Science [61]. ↵ Umetani K , Singer DH , McCraty R , Atkinson M . Twenty-four hour time domain heart rate variability and heart rate: relations to age and gender over nine decades . J Am Coll Cardiol . 1998 ; 31 : 593 – 601 . OpenUrl FREE Full Text [62]. ↵ Nowara EM , McDuff D , Veeraraghavan A. A Meta-Analysis of the Impact of Skin Type and Gender on Non-contact Photoplethysmography Measurements . 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2020 . p. 1148-55. [63]. ↵ Fine J , Branan KL , Rodriguez AJ , Boonya-Ananta T , Ajmal , Ramella-Roman JC , et al. Sources of Inaccuracy in Photoplethysmography for Continuous Cardiovascular Monitoring . Biosensors (Basel) . 2021 ; 11 : 126 . View the discussion thread. Back to top Previous Next Posted February 16, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Estimating heart rate variability using facial video photoplethysmography: a pilot validation study Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Estimating heart rate variability using facial video photoplethysmography: a pilot validation study Leszek Pstras , Tymoteusz Okupnik , Beata Ponikowska , Bartlomiej Paleczny medRxiv 2025.02.13.25322028; doi: https://doi.org/10.1101/2025.02.13.25322028 Share This Article: Copy Citation Tools Estimating heart rate variability using facial video photoplethysmography: a pilot validation study Leszek Pstras , Tymoteusz Okupnik , Beata Ponikowska , Bartlomiej Paleczny medRxiv 2025.02.13.25322028; doi: https://doi.org/10.1101/2025.02.13.25322028 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Cardiovascular Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (298) Cardiovascular Medicine (4421) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15215) Forensic Medicine (30) Gastroenterology (1122) Genetic and Genomic Medicine (6582) Geriatric Medicine (667) Health Economics (996) Health Informatics (4522) Health Policy (1366) Health Systems and Quality Improvement (1611) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15907) Intensive Care and Critical Care Medicine (1103) Medical Education (621) Medical Ethics (144) Nephrology (667) Neurology (6580) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3328) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5436) Public and Global Health (9215) Radiology and Imaging (2193) Rehabilitation Medicine and Physical Therapy (1368) Respiratory Medicine (1194) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (288) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff70008cf3e8e2e',t:'MTc3OTQwMjM1OQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();