RESOURCES

Key Publications & Clinical Studies

Read more about Happitech photoplethysmography (PPG) in clinical settings.

RESOURCES

Key Publications & Clinical Studies

Read more about Happitech photoplethysmography (PPG) in clinical settings.

PUBLICATIONS

Publications

01. Performance of an automated photoplethysmography-based artificial intelligence algorithm to detect atrial fibrillation

Mol D, Riezebos RK, Marquering HA, et al. Performance of an automated photoplethysmography-based artificial intelligence algorithm to detect atrial fibrillation. Cardiovascular Digital Health Journal. 2020;1(2):107-110.
doi:10.1016/j.cvdhj.2020.08.004

Backgrounds

Recently, an artificial intelligence smartphone-based PPG algorithm for detection of AF was developed by Happitech (Amsterdam, The Netherlands). The algorithm was trained using 2560 selected recordings retrieved from a worldwide online data donation campaign (Heart for Heart) and consists of 3 main components: (1) peak detection to measure R-R intervals; (2) quality; and (3) rhythm classification.

Objectives

To validate the performance of an automated photoplethysmography-based artificial intelligence algorithm to detect atrial fibrillation.

Country of Origin

NL

Patients

We validated the algorithm in patients with AF who were admitted to OLVG Hospital (Amsterdam, The Netherlands) for elective electrical cardioversion (ECV). PPG recordings were obtained directly before and after ECV using an iPhone 8 (Apple Inc., Cupertino, CA). Continuous electrocardiography was monitored simultaneously with the PPG heart rhythm recording for verification. The study was approved by the local medical ethics committee, and all participants provided written informed consent.

Methods

Steps taken by the photoplethysmography (PPG) algorithm to provide heart rhythm outcomes: First is detection of peaks using a shallow neural network; second is quality estimation using the support vector machine. After selection of 3 segments with the best quality in the third step, each segment is, based on rhythm features, classified as sinus rhythm (SR), atrial fibrillation (AF), or undetermined (UD). The final decision was made if ≥2 segments were classified in the same group.

Results

The shallow neural network showed excellent performance for peak detection. 98.1% sensitivity and specificity for detection of atrial fibrillation were obtained using a new automated plethysmography (PPG) algorithm. Predefined exclusion of recordings with low confidence boosted the diagnostic performance of the algorithm, resulting in 1.8% increase in sensitivity and 4.6% increase in specificity.

Comments and conclusions

The shallow neural network showed excellent performance for peak detection. 98.1% sensitivity and specificity for detection of atrial fibrillation were obtained using a new automated plethysmography (PPG) algorithm. Predefined exclusion of recordings with low confidence boosted the diagnostic performance of the algorithm, resulting in 1.8% increase in sensitivity and 4.6% increase in specificity.

02. Computationally efficient algorithm for photoplethysmography-based atrial fibrillation detection using smartphones

Schäck T, Safi Harb Y, Muma M, Zoubir AM. Computationally efficient algorithm for photoplethysmography-based atrial fibrillation detection using smartphones. IEEE. 2017:104-108. doi: 10.1109/EMBC.2017.8036773

Backgrounds

Atrial fibrillation (AF) is one of the major causes of stroke, heart failure, sudden death, and cardiovascular morbidity and the most common type of arrhythmia. Its diagnosis and the initiation of treatment, however, currently requires electrocardiogram (ECG)-based heart rhythm monitoring. The photoplethysmogram (PPG) offers an alternative method, which is convenient in terms of its recording and allows for self monitoring, thus relieving clinical staff and enabling early AF diagnosis.

Objectives

We propose an approach to acquire PPG signals from the video camera of a smartphone at a reduced computational cost and calculate a set of features to discriminate AF from NSR and to automatically exclude measurements with strong hand movements. With the use of feature selection and support vector machines (SVM) we achieve perfect detection of AF on the clinically recorded data.

Country of Origin

NL

Methods

PPG Signal Acquisition and Preprocessing: The PPG signals are acquired from the smartphones by using their camera and flash. In contrast to other studies that simply average 50 x 50 pixels of each video frame, we propose a novel method to better capture the variations of the pulsatile signal while reducing the computational cost and memory requirements.

Statistical Feature Extraction: Similar to other studies, we calculate statistical features from the PPG signal to distinguish between AF and normal sinus rhythm (NSR). In addition, we also make use of the features to detect strong hand movement during the measurement and label them as vibration (Vib). Implemented in a smartphone application, the user can be immediately informed that the measurement should be repeated. The statistical features are divided into two categories that are associated with the time-domain and the frequency domain of the PPG signal: Time-Domain Features and Frequency-Domain Features.

Feature Selection and Classification: We divide the classification procedure into two different stages: wrapper type feature selection and classification via support vector machines (SVM). Both stages are computational demanding but are not intended to be implemented in smartphone applications, as they only need the resulting classifications rules. First, the feature selection aims at finding the best feature combination of the presented set of features to distinguish between the classes. Feature selection reduces the computational cost and reveals the different levels of the features’ significance. Then, the SVM is applied to find the optimal decision equation to separate the classes. Finally, only the features of the best combination need to be calculated in the smartphone application and classification can be performed based on decision lines for pairs of features or hyperplanes in case of multiple features.

Results

Classification Results: The feature selection indicates that using only two features are sufficient for AF detection. Table I shows the results of the best feature pairs in terms of sensitivity, specificity and accuracy. The best single feature is the Shannon entropy of peak differences (ShE), which when combined with the median of peak rise height (mPRH) achieves a perfect classification accuracy of 100 %.

Comments and conclusions

In this paper, we proposed a PPG-based AF detection algorithm using the video camera of smartphones. The approach combines an enhanced PPG acquisition with a new set of discriminating features and a classification procedure that selects the most significant features and outputs decision equations for the discrimination between AF and NSR. The proposed method achieves perfect classification on a set of 326 measurements that were taken at a clinical prestudy. The low computational complexity allows for a mobile application that could be implemented in future.

03. (Upcoming) Validation of Photoplethysmography Using a Mobile Phone Application for the Assessment of Heart Rate Variability in the Context of HRV-Biofeedback

Van Dijk W, Huizink AC, Oosterman M, Lemmers-Jansen ILJ, de Vente W. Vrije Universiteit Amsterdam. 2022
To be published in 2022.

Objectives

Heart rate variability-biofeedback (HRV-BF) is an effective intervention to reduce stress and anxiety and requires accurate measures of realtime HRV. HRV can be measured through photoplethysmography (PPG) using the camera of a mobile phone. No studies have directly compared HRV-BF supported through PPG against classical electrocardiogram (ECG). Therefore, the current study aimed to validate PPG HRV measurements during HRV-BF against ECG.

Methods

A total of 57 healthy participants (70.69% female, age 17-60 years) received HRV-BF in the laboratory. Participants filled out questionnaires and performed five times a 5-min diaphragmatic breathing exercise at different paces (range: ~6.5 to ~4.5 breaths/
min). Four HRV-indices obtained through PPG and ECG were calculated and compared, that is, RMSSD, pNN50, LFpower, HFpower, as well as resonance frequency (optimal breathing pace).

Results

Overall, trivial-to-medium size differences between PPG-derived and ECG-derived RMSSD, pNN50, LFpower and HFpower were found. In addition, large and statistically significant correlations and ICC’s between PPG- and ECG-HRV indices were found. Furthermore, all Bland-Altman analyses (with just three incidental exceptions) showed good interchangeability of PPG- and ECGderived HRV indices. No systematic evidence for proportional bias was found for any of the HRV-indices. In addition, resonance frequency correspondence was good.

Comments and conclusions

PPG as calculated by Happitech-SDK is a highly appropriate method for the assessment of HRV indices relevant to perform HRV-BF. PPG is a promising replacement of ECG assessment to measure resonance frequency during HRV-BF.

04.Review of mobile applications for the detection and management of atrial fibrillation

Reading Turchioe M, Jimenez V, Isaac S, Alshalabi M, Slotwiner D, Masterson Creber R. Review of mobile applications for the detection and management of atrial fibrillation. Heart Rhythm O2. 2020;1(1):35-43. doi: 10.1016/j.hroo.2020.02.005.

Backgrounds

Free mobile applications (apps) that use photoplethysmography
(PPG) waveforms may extend
atrial fibrillation (AF) detection to underserved populations, but they
have not been rigorously evaluated.

Objectives

The purpose of this study was to systematically review and evaluate the quality, functionality, and adherence to self-management behaviors of existing mobile apps
for AF.

Country of Origin

NL

Methods

We systematically searched 3 app stores for apps that were free, available in English, and intended for use by patients to detect and manage AF. A minimum of 2 reviewers evaluated (1) app quality, using the Mobile Application Rating Scale (MARS); (2) functionality using published criteria; and (3) features that support 4 self-management behaviors (including PPG waveform monitoring) identified using evidence-based guidelines. Interrater reliability between the reviewers was calculated.

Results

Of 12 included apps, 5 (42%) scored above average for quality (MARS score ≥3.0). App quality was highest for their ease of use, navigation, layout, and visual appeal (eg, functionality and aesthetics) and lowest for their behavioral change support and subjective impressions of quality. The most common app functionalities were capturing and graphically displaying userentered data (n = 9 [75%]). Nearly all apps (n = 11 [92%]) supported PPG waveform monitoring, but only 2 (17%) supported all 4 selfmanagement behaviors. Interrater reliability was high (0.75 0.83).

Comments and conclusions

The reviewed apps had wide variability in quality, functionality, and adherence to self-management behaviors. Given the accessibility of these apps to underserved populations and the tremendous potential they hold for improving AF detection and management, high priority should be given to improving app quality and functionality.

05. Optimizing a Photophletysmography Algorhythm for Atrial Fibrillation Detection Using Crowdsourcing Research Data

Mol D, Safi Harb Y, Lobban TC, Riezebos RK, de Groot JR, de Jong JS. Optimizing a Photophletysmography Algorhythm for Atrial Fibrillation Detection Using Crowdsourcing Research Data. AHA Journals. 2018;138( Suppl_1). https://www.ahajournals.org/doi/abs/10.1161/circ.138.suppl_1.15538

Backgrounds

Photoplethysmography (PPG) records the reflection of light in blood perfused tissues, and can be used for heart rhythm analysis.
Advances in smartphone technology allow real time PPG recordings and analysis for heart rate. Atrial fibrillation (AF) is the most common arrhythmia. Early detection of AF can prevent strokes and their devastating sequels. Due to its irregularity proper adjudication of AF with PPG is dependent on signal quality and signal-to-noise ratio.

Objectives

To test the hypothesis that crowdsourcing is an effective method to collect a large number of PPG measurements to train an AF detection algorithm.

Country of Origin

NL

Methods

In 2017 the Heart for Heart campaign was initiated through a collaboration of cardiologists, Happitech, Arrhythmia alliance and Sudler-team. In an online campaign participants were asked to complete a 90 seconds PPG measurement and a questionnaire with demographic and arrhythmia history. Participants provided
informed consent through the Apple ResearchKit. Datasets were divided in 3 samples of 30 seconds. In these samples, signal quality was manually estimated based on waveform and vibrations. We used a 0-2 scoring scale per sample, a total score of >=5 was considered as high quality measurement. After classification, the data was used to train an AF algorithm.

Results

The Heart for Heart app was available from Jun ‘17 and was downloaded 9916 times within the first 3 months:
- 15256 datasets were collected from 74 countries worldwide.
- The questionnaire was completed by 12824 (84%) participants, of whom 7390 (59%) were male. Mean age was 45±16 years, and 2749 (21%) had a history of arrhythmias.
- A sample of 2560 datasets were manually assessed for signal quality. Recordings were of high quality in 2083 (81%) participants.
- Logistic regression analysis did not demonstrate age (OR 0.995, 95%CI 0.997-1.013) or BMI (OR 1.013, 95%CI 0.961-1.014) as predictors for low quality measurements.

Comments and conclusions

The online campaign proved to be effective and feasible to collect a large amount of worldwide research data. This study showed that the instructions and ease of use of the smartphone app were adequate to collect high quality data in 81% of users performing unsupervised measurements. Data quality was unrelated to age or BMI.

06. Prediction of vascular ageing based on smartphone acquired PPG signals

Dall’Olio L, Curti N, Remondini D, et al. Prediction of vascular aging based on smartphone acquired PPG signals. Sci Rep. 2020;10(1):19756 doi: 10.1038/s41598-020-76816-6

Backgrounds

Photoplethysmography (PPG) measured by smartphone has the
potential for a large scale, noninvasive, and easy-to-use screening tool. Vascular aging, in particular, is characterized by a gradual change of the vascular structure and function, and increasing arterial stiffness is considered to be the hallmark of vascular aging. Arterial stiffness can be measured by pulse wave velocity (PWV), or by the use
of the PPG technique. In particular, some aging indexes (AGI) can be calculated from the second
derivative of the PPG (SDPPG) waveform.

Objectives

To investigate the feasibility of using PPG to predict healthy vascular aging (HVA) based on two approaches: machine learning (ML) and deep learning (DL). Without having the heavy computational cost of DL, ML might be advantageous in finding potential biomarkers for HVA prediction.

Country of Origin

NL

Patients

Our motivating crowd-sourced data comes from the Heart for Heart (H4H) initiative, promoted by the Arrhythmia Alliance, the Atrial Fibrillation Association, Happitech and other partners. The aim of H4H is to gather millions of cardiac measurements and to increase the pace of progress on AF diagnostic technology.
The primary advantages of using this population cohort data are: abundance of PPG recordings in large samples (ca. 10,000) and relatively long sequences (90 seconds), and free access to raw
PPG signals via Happitech app

Methods

The PPG is preprocessed by computing a centered moving average (CMA) and subtracting it from the raw signal (to perform a high-pass filter). Then feature extraction is applied to PPG signal for a machine learning approach using both Ride Regression and Convolution Neural Networks.

Results

To validate the results, the independent test set was used . The AUC obtained using only one feature (i.e., a, slope-AC, or tpr) was around 0.8 . The following models are considered: (i) covariates (weight, height, sex, smoking), (ii) the best two PPG features (a and tpr) from ML, (iii) covariates and these two features, (iv) th  best performing CNN, and (v) covariates and all PPG features. One should take note that the model (v) is not recommended due to the overfitting, which may not lead to such a good performance on a different dataset. For comparing the prediction performance of the five models, the area under the ROC curve (AUC) was computed. By adding the PPG features (tpr and a) to the covariates, AUC increased from 0.742 to 0.947. The 12-layer ResNet model (AUC=0.953) performed similarly to including all variables (AUC=0.954).

Comments and conclusions

We showed that PPG measured by smartphone has the potential for large scale, non-invasive, patient-led screening. This work contributes to establishing a generally accepted algorithm based on open data and software, which is of major importance to reproduce the procedures, and to further improve and develop methods.

07. Stress- and smoke free pregnancy study protocol: a randomized controlled trial of a personalized eHealth intervention including heart rate variability biofeedback to support pregnant women quit smoking via stress reduction

van Dijk W, Oosterman M, Jansen I, de Vente W, Huizink A. Stress- and smoke free pregnancy study protocol: a randomized controlled trial of a personalized eHealth intervention including heart rate variability-biofeedback to support pregnant women quit smoking via stress reduction. BMC Public Health. 2021;21(1). doi:10.1186/s12889-021-10910-w

Backgrounds

Maternal smoking and stress during pregnancy are associated with adverse health effects for women themselves and are risk factors for adverse developmental outcomes of the unborn child. Smoking and stress seem to be intertwined in various ways. First, the majority of smoking pregnant women are of lower socio-economic status, which is associated with higher levels of perceived stress. Second, smoking women often report they smoke because they feel stressed. Third, quitting smoking often increases perceived stress levels initially.
Therefore, effective interventions are needed to support women with smoking cessation by reducing stress.

Objectives

The aim of this study is to test
the effectiveness of an eHealth
intervention on stress reduction and smoking cessation.

Country of Origin

NL

Study Type

This study is a two-arm, parallel, single blinded randomized controlled trial (RCT) including four measurement waves: preintervention or baseline (t0), post-intervention (t1), two weeks following birth (t2) and three months after birth (t3). Pregnant, smoking women will be randomized to one of two groups: 1) personalized eHealth intervention (“Together with Eva”) on the smartphone followed for 8 weeks; or 2) active control condition which includes psycho-education on stress, smoking, and pregnancy through a website.

Patients

The Stress- and Smoke Free Start of Life (SSFSL) study is a randomized controlled trial (RCT) comparing a personalized eHealth intervention with a control condition. Inclusion criteria for the women are: (1) > 18 years of age, (2) < 28 weeks pregnant at recruitment, (3) currently smoking

Methods

Consenting participants will be randomly assigned to the intervention or control group. Participants allocated to the intervention group will receive an 8-week intervention delivered on their smartphone. The application includes psycho-education on pregnancy, stress, and smoking (cessation); stress-management training consisting of Heart Rate Variability-biofeedback; and a personalized stop-smoking-plan. Participants in the control condition will be invited to visit a webpage with information on - 12 - pregnancy, stress, and smoking (cessation). Study outcomes will be collected via online questionnaires, at four timepoints: preintervention (baseline; t0), post-intervention (8 weeks + 1 day after t0; t1), follow up at two weeks after birth (t2), and follow up at three months after birth (t3).

Outcome Measures

The primary outcome measure is self-reported smoking cessation. Secondary outcomes include daily self-reported number of cigarettes smoked, perceived stress, pregnancy experience, birth outcomes, and negative affectivity scores of the baby. Moreover, the mediating effect of stress reduction on smoking cessation will be examined, and possible moderators will be tested

Results

After receiving permission by the Medical Ethics Committee to start with the inclusion of participants, the first participants were included in July 2020. Currently, data collection is ongoing. The main results are expected to be published in 2022.

08. Quantitative analysis of smartphone PPG data for heart monitoring

Bussola F. Quantitative analysis of smartphone PPG data for heart monitoring
Alma Mater Studiorum - Universita di Bologna. 2017/2018. https://core.ac.uk/download/pdf/211577183.pdf

Backgrounds

Heart for Heart is a crowd-sourced initiative promoted by the Arrhythmia Alliance, the Atrial Fibrillation Association, Happitech and other partners, which aims at gathering a million of cardiac measurements [9]. It also aims at increasing awareness of atrial fibrillation and accelerating the pace of progress on atrial fibrillation diagnostic technology. Happitech has conducted clinical trials of its technology in the past in collaboration with hospitals in Amsterdam (The Netherlands) and is now (as of time of writing, in 2018 and 2019) conducting more trials involving also UMC Utrecht.

Objectives

1. characterizing the data available;

2. identifying possible techniques suitable to classify two different rhythms in data,

3. normal sinus rhythm (NSR) and atrial fibrillation (Afib);

4. analysing the issue of signal quality

Country of Origin

NL

Methods

We aim at characterizing a dataset of 1572 subjects, whose signals have been crowdsourced by collecting measurements via a dedicated smartphone app, using the embedded camera. We evaluate the distributions of three features of our signals: the peak area, amplitude and the time interval between two successive pulses. We evaluate if some factors affected the distributions, discovering that the strongest effects are for age and BMI groupings. We evaluate the results agreement between the R G B channels of acquisition, finding good agreement between the first two.

After finding signal quality indexes in literature, we use a subset of them in a classification task, combined with dynamic time warping distance, a technique that matches a signal to a template. We achieve an accuracy of 89% on the test set, for binary quality classification.

On the chaotic temporal series we evaluate the appearance of different types of rhythms on Poincaré plots and we quantify the results by a measure of their 3D spread. We perform this on a set of 20 subjects, 10 NSR and 10 Afib, finding significant differences between their 3D morphologies. We extend our analysis to the larger dataset, obtaining some significant results.

Comments and conclusion

The promising and vast field of PPG-based cardiac health monitoring faces great challenges if it is to provide real-time feedback and diagnostics to end users or health care providers. The open issue of dealing with noisy measurements when addressing different type of rhythms for clinical purposes is such, open.

Our research points to the direction of the strong need for accurately labelled measurements and multivariate analysis, to conclusively answer the questions of underlying effects in different groups. A heart rhythm classification that takes into account also signal quality seems to be within reach, especially by exploiting the power of machine learning data analysis techniques.

The physical approach of recurrence of the chaotic time series yields promising findings as far as discerning between NSR and Afib rhythms is concerned. At this stage, classification is potentially possible via analysing the 3D spread of the Poincaré plots of the time intervals of a signal.

09. Opportunistic screening for atrial fibrillation using a photoplethysmography technique in geriatric patients, a preliminary analysis of the Dutch-GERAF Study

Mr Zwart LAR; Doctor Jansen RWMM; Miss Spruit JR; Doctor Pisters R; Doctor Riezebos RK; Professor De Groot JR; Doctor Hemels MEW - EP Europace, Volume 25, Issue Supplement_1, June 2023, euad122.554

Backgrounds

Undiagnosed atrial fibrillation (AF) poses a great risk for stroke in highrisk patients. Different screening strategies have been developed to identify these patients. Geriatric patients often have multimorbidity and are to be considered a population at very high risk for the development of AF and cardiovascular events, but may have less access to wearable screening devices.

Objectives

The Dutch multicentre study into opportunistic screening of Geriatric patients for Atrial Fibrillation using a PPG smartphone App (the DutchGERAF Study) applies opportunistic screening to ambulatory geriatric patients, to identify new cases of AF and optimize stroke prevention. This preliminary analysis describes the feasibility and willingness to use a smartphone photoplethysmography (PPG) application in an ambulatory cohort.

Country of Origin

NL

Methods

Analysis of participation data and PPG use, of all included patients aged 70 years or older within the GERAF study. The GERAF study is a multicentre study on opportunistic screening for AF using a PPG application among patients visiting a geriatric outpatient clinic. All patients without a pacemaker/ICD, severe tremor, or severe dementia were invited to participate, during a 6 months screening period. Patients were stimulated to use the application at home on their own mobile device at least 3 times within 6 months following inclusion. If preferred by the patient, the application was installed on the mobile device of a spouse or other family member. The PPG signals were analysed using the commercially available algorithm.

Results

Data on PPG use for patients was available for 568 patients. Of these, 122 (21.5%) had a history of AF, and therefore not included in this analysis. Data on PPG use was analysed of 446 patients, 240 were female (54%), the average age was 79 ± 5.5 years, the oldest patient was 95 years old. In total, patients performed 2232 PPG recordings within 6 months, in which a sufficient signal was acquired in 1874 (84%) recordings. The signal appeared sufficient at the first attempt in 79.3% of the recordings. In case of a failed first attempt, 60.5% of the patients immediately performed at least one extra recording until sufficient signal quality was achieved, and overall 61.2% of the patients performed at least 2 PPG recordings. Only 69 (15.5%) patients who consented to participation, did not perform any recordings.

Conclusion

Conclusion: This analysis of preliminary GERAF PPG data shows that very old subjects (with a mean age of almost 80 years) are willing and able to use eHealth applications on a mobile device. Most PPG recordings were of sufficient signal quality, and more than half of the patients were willing to perform repeated recordings.

Our Research Partners

We have multiple research partnerships with leading hospitals and academic institutions including Mount Sinai, Boston Children’s Hospital, OVLG, Vrije Universiteit Amsterdam, Hartstichting, and UMC Utrecht.

Scroll to Top