Diagnostic assessment of a deep learning system for detecting atrial fibrillation in pulse waveforms

Analysis of Relevant Features from Photoplethysmographic Signals for Atrial Fibrillation Classification

Atrial Fibrillation (AF) is the most common cardiac arrhythmia found in clinical practice. It affects an estimated 33.5 million people, representing approximately 0.5% of the world's population. Electrocardiogram (ECG) is the main diagnostic criterion for AF. Recently, photoplethysmography (PPG) has emerged as a simple and portable alternative for AF detection. However, it is not completely clear which are the most important features of the PPG signal to perform this process. The objective of this paper is to determine which are the most relevant features for PPG signal analysis in the detection of AF. This study is divided into two stages: (a) a systematic review carried out following the Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies (PRISMA-DTA) statement in six databases, in order to identify the features of the PPG signal reported in the literature for the detection of AF, and (b) an experimental evaluation of them, using machine learning, in order to determine which have the greatest influence on the process of detecting AF. Forty-four features were found when analyzing the signal in the time, frequency, or time-frequency domains. From those 44 features, 27 were implemented, and through machine learning, it was found that only 11 are relevant in the detection process. An algorithm was developed for the detection of AF based on these 11 features, which obtained an optimal performance in terms of sensitivity (98.43%), specificity (99.52%), and accuracy (98.97%).

Photoplethysmography based atrial fibrillation detection: a review

Atrial fibrillation (AF) is a cardiac rhythm disorder associated with increased morbidity and mortality. It is the leading risk factor for cardioembolic stroke and its early detection is crucial in both primary and secondary stroke prevention. Continuous monitoring of cardiac rhythm is today possible thanks to consumer-grade wearable devices, enabling transformative diagnostic and patient management tools. Such monitoring is possible using low-cost easy-to-implement optical sensors that today equip the majority of wearables. These sensors record blood volume variations-a technology known as photoplethysmography (PPG)-from which the heart rate and other physiological parameters can be extracted to inform about user activity, fitness, sleep, and health. Recently, new wearable devices were introduced as being capable of AF detection, evidenced by large prospective trials in some cases. Such devices would allow for early screening of AF and initiation of therapy to prevent stroke. This review is a summary of a body of work on AF detection using PPG. A thorough account of the signal processing, machine learning, and deep learning approaches used in these studies is presented, followed by a discussion of their limitations and challenges towards clinical applications.

Validation of Single Centre Pre-Mobile Atrial Fibrillation Apps for Continuous Monitoring of Atrial Fibrillation in a Real-World Setting: Pilot Cohort Study

Background

Atrial fibrillation is the most common recurrent arrhythmia in clinical practice, with most clinical events occurring outside the hospital. Low detection and nonadherence to guidelines are the primary obstacles to atrial fibrillation management. Photoplethysmography is a novel technology developed for atrial fibrillation screening. However, there has been limited validation of photoplethysmography-based smart devices for the detection of atrial fibrillation and its underlying clinical factors impacting detection.

Objective

This study aimed to explore the feasibility of photoplethysmography-based smart devices for the detection of atrial fibrillation in real-world settings.

Methods

Subjects aged ≥18 years (n=361) were recruited from September 14 to October 16, 2018, for screening of atrial fibrillation with active measurement, initiated by the users, using photoplethysmography-based smart wearable devices (ie, a smart band or smart watches). Of these, 200 subjects were also automatically and periodically monitored for 14 days with a smart band. The baseline diagnosis of “suspected” atrial fibrillation was confirmed by electrocardiogram and physical examination. The sensitivity and accuracy of photoplethysmography-based smart devices for monitoring atrial fibrillation were evaluated.

Results

A total of 2353 active measurement signals and 23,864 periodic measurement signals were recorded. Eleven subjects were confirmed to have persistent atrial fibrillation, and 20 were confirmed to have paroxysmal atrial fibrillation. Smart devices demonstrated >91% predictive ability for atrial fibrillation. The sensitivity and specificity of devices in detecting atrial fibrillation among active recording of the 361 subjects were 100% and about 99%, respectively. For subjects with persistent atrial fibrillation, 127 (97.0%) active measurements and 2240 (99.2%) periodic measurements were identified as atrial fibrillation by the algorithm. For subjects with paroxysmal atrial fibrillation, 36 (17%) active measurements and 717 (19.8%) periodic measurements were identified as atrial fibrillation by the algorithm. All persistent atrial fibrillation cases could be detected as “atrial fibrillation episodes” by the photoplethysmography algorithm on the first monitoring day, while 14 (70%) patients with paroxysmal atrial fibrillation demonstrated “atrial fibrillation episodes” within the first 6 days. The average time to detect paroxysmal atrial fibrillation was 2 days (interquartile range: 1.25-5.75) by active measurement and 1 day (interquartile range: 1.00-2.00) by periodic measurement (P=.10). The first detection time of atrial fibrillation burden of <50% per 24 hours was 4 days by active measurement and 2 days by periodic measurementThe first detection time of atrial fibrillation burden of >50% per 24 hours was 1 day for both active and periodic measurements (active measurement: P=.02, periodic measurement: P=.03).

Conclusions

Photoplethysmography-based smart devices demonstrated good atrial fibrillation predictive ability in both active and periodic measurements. However, atrial fibrillation type could impact detection, resulting in increased monitoring time.

Trial Registration

Chinese Clinical Trial Registry of the International Clinical Trials Registry Platform of the World Health Organization ChiCTR-OOC-17014138; http://www.chictr.org.cn/showprojen.aspx?proj=24191.

Smartwatch for the Detection of Atrial Fibrillation

Atrial fibrillation (AF) is the most common arrhythmia and poses a substantial economic burden due to associated thromboembolic complications. Screening for AF may theoretically be effective, but there is no consensus regarding the optimal screening method because the available tools are either invasive or not cost-effective. Recently, smartwatch industry has received a surge of interest for this purpose by introducing technologies such as photoplethysmography, artificial intelligence, and actual electrodes taking an electrocardiogram to measure and analyze heart rate and rhythm with relatively acceptable accuracy. Combined with other features such as ease of use and connectivity, smartwatches can potentially be used for large-scale AF screening and might eventually replace the current gold standards. In this review, we discuss the feasibility of this approach and summarize the current evidence on AF detection with smartwatches.

Detecting Atrial Fibrillation and Atrial Flutter in Daily Life Using Photoplethysmography Data

OBJECTIVE

Photoplethysmography (PPG) enables unobtrusive heart rate monitoring, which can be used in wrist-worn applications. Its potential for detecting atrial fibrillation (AF) has been recently presented. Besides AF, another cardiac arrhythmia increasing stroke risk and requiring treatment is atrial flutter (AFL). Currently, the knowledge about AFL detection with PPG is limited. The objective of our study was to develop a model that classifies AF, AFL, and sinus rhythm with or without premature beats from PPG and acceleration data measured at the wrist in daily life.

METHODS

A dataset of 40 patients was collected by measuring PPG and accelerometer data, as well as electrocardiogram as a reference, during 24-hour monitoring. The dataset was split into 75%-25% for training and testing a Random Forest (RF) model, which combines features from PPG, inter-pulse intervals (IPI), and accelerometer data, to classify AF, AFL, and other rhythms. The performance was compared to an AF detection algorithm combining traditional IPI features for determining the robustness of the accuracy in presence of AFL.

RESULTS

The RF model classified AF/AFL/other with sensitivity and specificity of 97.6/84.5/98.1% and 98.2/99.7/92.8%, respectively. The results with the IPI-based AF classifier showed that the majority of false detections were caused by AFL.

CONCLUSION

The PPG signal contains information to classify AFL in the presence of AF, sinus rhythm, or sinus rhythm with premature contractions.

SIGNIFICANCE

PPG could indicate presence of AFL, not only AF.

Mobile Photoplethysmographic Technology to Detect Atrial Fibrillation

BACKGROUND

Low detection and nonadherence are major problems in current management approaches for patients with suspected atrial fibrillation (AF). Mobile health devices may enable earlier AF detection and improved AF management.

OBJECTIVES

This study sought to investigate the effectiveness of AF screening in a large population-based cohort using smart device-based photoplethysmography (PPG) technology, combined with a clinical care AF management pathway using a mobile health approach.

METHODS

AF screening was performed with smart devices using PPG technology, which were made available for the population ≥18 years of age across China. Monitoring for at least 14 days with a wristband (Honor Band 4) or wristwatch (Huawei Watch GT, Honor Watch, Huawei Technologies Co., Ltd., Shenzhen, China) was allowed. The patients with "possible AF" episodes using the PPG algorithm were further confirmed by health providers among the MAFA (mobile AF app) Telecare center and network hospitals, with clinical evaluation, electrocardiogram, or 24-h Holter monitoring.

RESULTS

There were 246,541 individuals who downloaded the PPG screening app, and 187,912 individuals used smart devices to monitor their pulse rhythm between October 26, 2018, and May 20, 2019. Among those with PPG monitoring (mean age 35 years, 86.7% male), 424 (of 187,912, 0.23%) (mean age 54 years, 87.0% male) received a "suspected AF" notification. Of those effectively followed up, 227 individuals (of 262, 87.0%) were confirmed as having AF, with the positive predictive value of PPG signals being 91.6% (95% confidential interval [CI]: 91.5% to 91.8%). Both suspected AF and identified AF markedly increased with age (p for trend <0.001), and individuals in Northeast China had the highest proportion of detected AF of 0.28% (95% CI: 0.20% to 0.39%). Of the individuals with identified AF, 216 (of 227, 95.1%) subsequently entered a program of integrated AF management using a mobile AF application; approximately 80% of high-risk patients were successfully anticoagulated.

CONCLUSIONS

Based on the present study, continuous home monitoring with smart device-based PPG technology could be a feasible approach for AF screening. This would help efforts at screening and detection of AF, as well as early interventions to reduce stroke and other AF-related complications. (Mobile Health [mHealth] Technology for Improved Screening, Patient Involvement and Optimizing Integrated Care in Atrial Fibrillation [MAFA II]; ChiCTR-OOC-17014138).

Deep Learning Approaches to Detect Atrial Fibrillation Using Photoplethysmographic Signals: Algorithms Development Study

Background

Wearable devices have evolved as screening tools for atrial fibrillation (AF). A photoplethysmographic (PPG) AF detection algorithm was developed and applied to a convenient smartphone-based device with good accuracy. However, patients with paroxysmal AF frequently exhibit premature atrial complexes (PACs), which result in poor unmanned AF detection, mainly because of rule-based or handcrafted machine learning techniques that are limited in terms of diagnostic accuracy and reliability.

Objective

This study aimed to develop deep learning (DL) classifiers using PPG data to detect AF from the sinus rhythm (SR) in the presence of PACs after successful cardioversion.

Methods

We examined 75 patients with AF who underwent successful elective direct-current cardioversion (DCC). Electrocardiogram and pulse oximetry data over a 15-min period were obtained before and after DCC and labeled as AF or SR. A 1-dimensional convolutional neural network (1D-CNN) and recurrent neural network (RNN) were chosen as the 2 DL architectures. The PAC indicator estimated the burden of PACs on the PPG dataset. We defined a metric called the confidence level (CL) of AF or SR diagnosis and compared the CLs of true and false diagnoses. We also compared the diagnostic performance of 1D-CNN and RNN with previously developed AF detectors (support vector machine with root-mean-square of successive difference of RR intervals and Shannon entropy, autocorrelation, and ensemble by combining 2 previous methods) using 10 5-fold cross-validation processes.

Results

Among the 14,298 training samples containing PPG data, 7157 samples were obtained during the post-DCC period. The PAC indicator estimated 29.79% (2132/7157) of post-DCC samples had PACs. The diagnostic accuracy of AF versus SR was 99.32% (70,925/71,410) versus 95.85% (68,602/71,570) in 1D-CNN and 98.27% (70,176/71,410) versus 96.04% (68,736/71,570) in RNN methods. The area under receiver operating characteristic curves of the 2 DL classifiers was 0.998 (95% CI 0.995-1.000) for 1D-CNN and 0.996 (95% CI 0.993-0.998) for RNN, which were significantly higher than other AF detectors (P<.001). If we assumed that the dataset could emulate a sufficient number of patients in training, both DL classifiers improved their diagnostic performances even further especially for the samples with a high burden of PACs. The average CLs for true versus false classification were 98.56% versus 78.75% for 1D-CNN and 98.37% versus 82.57% for RNN (P<.001 for all cases).

Conclusions

New DL classifiers could detect AF using PPG monitoring signals with high diagnostic accuracy even with frequent PACs and could outperform previously developed AF detectors. Although diagnostic performance decreased as the burden of PACs increased, performance improved when samples from more patients were trained. Moreover, the reliability of the diagnosis could be indicated by the CL. Wearable devices sensing PPG signals with DL classifiers should be validated as tools to screen for AF.

Connected Health Technology for Cardiovascular Disease Prevention and Management

PURPOSE OF THE REVIEW

Advances in computing power and wireless technologies have reshaped our approach to patient monitoring. Medical grade sensors and apps that were once restricted to hospitals and specialized clinic are now widely available. Here, we review the current evidence supporting the use of connected health technologies for the prevention and management of cardiovascular disease in an effort to highlight gaps and future opportunities for innovation.

RECENT FINDINGS

Initial studies in connected health for cardiovascular disease prevention and management focused primarily on activity tracking and blood pressure monitoring but have since expanded to include a full panoply of novel sensors and pioneering smartphone apps with targeted interventions in diet, lipid management and risk assessment, smoking cessation, cardiac rehabilitation, heart failure, and arrhythmias. While outfitting patients with sensors and devices alone is infrequently a lasting solution, monitoring programs that include personalized insights based on patient-level data are more likely to lead to improved outcomes. Advances in this space have been driven by patients and researchers while healthcare systems remain slow to fully integrate and adequately adapt these new technologies into their workflows. Cardiovascular disease prevention and management continue to be key focus areas for clinicians and researchers in the connected health space. Exciting progress has been made though studies continue to suffer from small sample size and limited follow-up. Efforts that combine home patient monitoring, engagement, and personalized feedback are the most promising. Ultimately, combining patient-level ambulatory sensor data, electronic health records, and genomics using machine learning analytics will bring precision medicine closer to reality.

Deep Learning in Cardiology

The medical field is creating large amount of data that physicians are unable to decipher and use efficiently. Moreover, rule-based expert systems are inefficient in solving complicated medical tasks or for creating insights using big data. Deep learning has emerged as a more accurate and effective technology in a wide range of medical problems such as diagnosis, prediction, and intervention. Deep learning is a representation learning method that consists of layers that transform data nonlinearly, thus, revealing hierarchical relationships and structures. In this review, we survey deep learning application papers that use structured data, and signal and imaging modalities from cardiology. We discuss the advantages and limitations of applying deep learning in cardiology that also apply in medicine in general, while proposing certain directions as the most viable for clinical use.