Necklace‐embedded electrocardiogram for the detection and diagnosis of atrial fibrillation

Abstract Background Atrial fibrillation (AF) is the major cause of stroke since approximately 25% of all strokes are of cardioembolic‐origin. The detection and diagnosis of AF are often challenging due to the asymptomatic and intermittent nature of AF. Hypothesis A wearable electrocardiogram (ECG)‐device could increase the likelihood of AF detection. The aim of this study was to evaluate the feasibility and reliability of a novel, consumer‐grade, single‐lead ECG recording device (Necklace‐ECG) for screening, identifying and diagnosing of AF both by a cardiologist and automated AF‐detection algorithms. Methods A thirty‐second ECG was recorded with the Necklace‐ECG device from two positions; between the palms (palm) and between the palm and the chest (chest). Simultaneously registered 3‐lead ECGs (Holter) served as a golden standard for the final rhythm diagnosis. Two cardiologists interpreted independently in a blinded fashion the Necklace‐ECG recordings from 145 patients (66 AF and 79 sinus rhythm, SR). In addition, the Necklace‐ECG recordings were analyzed with an automatic AF detection algorithm. Results Two cardiologists diagnosed the correct rhythm of the interpretable Necklace‐ECG with a mean sensitivity of 97.2% and 99.1% (palm and chest, respectively) and specificity of 100% and 98.5%. The automatic arrhythmia algorithm detected the correct rhythm with a sensitivity of 94.7% and 98.3% (palm and chest) and specificity of 100% of the interpretable measurements. Conclusions The novel Necklace‐ECG device is able to detect AF with high sensitivity and specificity as evaluated both by cardiologists and an automated AF‐detection algorithm. Thus, the wearable Necklace‐ECG is a new, promising method for AF screening. Clinical trial registration: Study was registered in the ClinicalTrials.gov database (NCT03753139).


Conclusions:
The novel Necklace-ECG device is able to detect AF with high sensitivity and specificity as evaluated both by cardiologists and an automated AF-detection algorithm. Thus, the wearable Necklace-ECG is a new, promising method for AF screening. Clinical trial registration: Study was registered in the ClinicalTrials.gov database (NCT03753139).

K E Y W O R D S
Awario analysis service, arrhythmia, atrial fibrillation, ECG, stroke, Suunto Movesense

| INTRODUCTION
Atrial fibrillation (AF) is a rapidly growing global public health problem. 1 At the population level, the lifetime risk for developing AF is around 25%. 2 One in four middle-aged adults in the EU and the United States will develop AF. 3,4 The most severe complication of AF is an embolic stroke; approximately 25% of all strokes are caused by AF. 5 Stroke imposes a significant financial burden on the health care system, with a total cost of over €40 billion in both the European Union and the United States. 6,7 These costs will continue to rise as the population ages and lives longer. It has been claimed that two out every three AF related strokes can be prevented by the provision of appropriate oral anticoagulation therapy as long as AF is diagnosed early enough. 8 New screening methods for AF in different scenarios are being intensively studied. 9,10,11 Current guidelines state that a single-lead electrocardiogram (ECG) tracing of ≥30 s or 12-lead ECG analyzed by a physician is necessary to establish a definitive diagnosis of AF. 12 Traditional invasive and non-invasive methods for AF detection, such as ambulatory ECGs (Holter), loop recorders and prolonged ambulatory ECGs (mobile cardiovascular telemetry) are expensive and require interpretation by healthcare professionals. 13 There are several commercial handheld ECG devices for AF screening, designed for consumer use. [14][15][16][17][18][19] These kinds of devices can represent a relatively cost-efficient solution for AF screening, since many patients are motivated to monitor their own health. 15,17 However, to be effective, users must incorporate such devices into their everyday use. It is evident that device design plays a significant role in user adoption and therefore clever design can increase the usage of these devices over the long term. 20 Many commercial health products can be too technical and complicated for the consumers in high-risk groups, such as the elderly. 21 Therefore, in this study, we evaluated the feasibility and accuracy of a novel ECG recording technique, a wearable Necklace-ECG, for the detection of AF by (a) cardiologists and (b) an automated algorithm.

| Study design
The study was performed as a single-center case-control study. Local Ethics Committee approved the study protocol (01/08/2018) and the study was registered in the ClinicalTrials.gov database (NCT03753139).

| Study population
A total of 260 patients was screened in our emergency department between November 2018-May 2019. The flowchart of the screening of study participants is presented in Figure 1

| Necklace-ECG recording
All study subjects performed an at least thirty-second self-performed ECG recordings in the sitting, half-sitting or lying down positions using two measurement positions with a single-lead Necklace-embedded ECG recorder (Including Movesense ECG-sensor, Suunto, Vantaa, Finland, Necklace-ECG, Figure 2 and Figure S1). First, the subjects were holding the recorder between the palms of their hands (palm) and secondly, between the chest and the right palm (chest), both measurement positions simulate lead I ECG. A maximum of three measuring attempts per position were allowed.

| Continuous ECG recording
The final rhythm classification was based on simultaneously recorded 3-lead ECG (Holter, Faros 360, Bittium, Oulu, Finland) in the sitting, half-sitting or lying down positions. Two experienced cardiologists interpreted the ECG recordings independently. In case of disagreement, they reviewed the ECG recording together to confirm the final rhythm analysis.
The clinical characteristics of the patients including age, gender, height, weight, BMI, concomitant diseases and medications were collected by interview and confirmed/complemented from the medical records.

| Necklace-ECG analysis
Two experienced cardiologists interpreted the Necklace-ECG recordings independently in a random order and blinded to the Holter ECG recordings and the initial 12-lead ECG. In case of disagreement, they reviewed the ECG recording together to confirm the final rhythm analysis. They divided the rhythm of the Necklace-ECG recordings into four categories: sinus rhythm, atrial fibrillation, other and noninterpretable. Accordingly, the quality of Necklace-ECG was classified into four categories: high, average, poor and failure. In addition, the possibility of detecting P-waves from the ECG strips showing SR (yes/no) was assessed.
A commercial arrhythmia analysis service (Awario, Heart2Save, Kuopio, Finland) was used to test the accuracy of automatic AF-screening from the Necklace-ECG recording. The Necklace-ECG data was transferred to the cloud-based analysis service via a mobile phone application ( Figure 2). An automatic arrhythmia algorithm classified the Necklace-ECG strips into three categories: sinus rhythm, atrial fibrillation or non-interpretable.

| Statistical analysis
The sample size was estimated to be 150 observations with an assumed sensitivity of 95% with the method and with a 3.5% margin error, such that the sensitivity parameter with 95% certainty lay between 91.5% and 98.5%. The AF and SR groups were compared using t-test for continuous variables and χ 2 test or Fisher's exact tests for dichotomous variables. The Kappa-coefficient was calculated to measure the level of consensus between the "golden standard" (rhythm diagnosis from the Holter ECG recordings) and the cardiologists AF-diagnosis and algorithm AF-detections from the Necklace-ECG. Additionally, the sensitivity and specificity of AF detection were determined for the cardiologists and automatic arrhythmia detection algorithm. All significance tests were two-tailed and p ≤ .05 was considered statistically significant. The data was analyzed using IBM SPSS statistics software version 25.

| Clinical characteristics
The consensus rhythm of two experienced cardiologists from 3-lead Holter ECG was used as "golden standard" for the rhythm analysis.
F I G U R E 2 Schematic presentation of Necklace-electrocardiogram enabled automatic arrhythmia analysis AF patients were older than the SR patients (72.7 ± 14.1 vs. 61.5 ± 18.1 years, p < .001). AF patients also presented more often with a history of paroxysmal AF (p < .001), asymptomatic AF (p < .001), congestive heart failure (p < .001) and were more often being treated with anticoagulation (p < .001), beta-blocker (p < .001) and digoxin (p < 0.05) therapy. AF patients also reported more often respiratory distress (p < .001), compared to the SR group.  Figure S2.

| Diagnosing AF by Necklace-ECG
The consensus of the two cardiologists regarding AF-diagnosis from the Holter and Necklace ECGs are presented in Table 2. Based on the interpretable Necklace-ECG recordings, AF was diagnosed with a mean sensitivity of 97.2% using the palm and 99.1% using the chest measurements. The mean specificity was 100% and 98.5% using the palm and the chest measurements. For the cardiologists interpretation of the Necklace-ECG, the mean kappa's coefficient with the cardiologists diagnosis from Holter ECG was almost perfect (κ = 0.975) using both the palm and the chest measurements, see

| AF-screening with an automatic algorithm
The automatic arrhythmia algorithm found 93.1% and 92.4% (palm and chest, respectively) of ECG recordings interpretable. The AF detection sensitivity was 94.7% (palm) 98.3% (chest) and specificity  Table 3.

| DISCUSSION
The main finding of our study was that the wearable Necklace-ECG device was able to produce ECG recordings with high quality for the detection and diagnosis of AF and SR with high sensitivity and specificity as evaluated by two cardiologists as well as by an automated arrhythmia detection algorithm.  13 However, these methods are invasive and expensive and thus, they are not suitable for large scale AF screening. 13 Therefore, novel methods for AF detection are needed, especially as the incidence of AF will increase due to the aging of populations. One cost-effective solution for AF screening could be an easily accessible "handheld ECG" device designed for consumer use, such as the Necklace-ECG [14][15][16][17] ECG-based handheld devices have been evaluated for AF-detection, either interpreted by physicians or by using automated algorithms. [14][15][16][17] In these studies, the handheld devices have been compared with the traditional 12-or 6-lead ECG registration with inconclusive results. In terms of signal quality, between 0.8% to 13% of handheld ECG recordings were reported to be inadequate for rhythm diagnosis as interpreted by physicians 14,15,17,22,23 and 3.9%-33.7% as uncapable of being interpreted by an automated algorithm. 14,17,22,23 Our study is in line with these earlier reports; Necklace-ECG recordings were judged as non-interpretable in 9.0%-23% of all patients and in 16%-18% of patients with AF. Somewhat surprisingly, the quality of ECG recordings was sufficient for AF detection more often when interpreted by the automated algorithm than when interpreted by the cardiologist. Only 6%-8% of the Necklace-ECG recording were deemed non-interpretable with the automatic arrhythmia algorithm. One reason for the inadequate quality was that in cases where there was any uncertainty, the cardiologists were advised to label the recordings as non-interpretable.
In our study, the Necklace-ECG yielded a sensitivity of 95%-100% and specificity of 98%-100% with either physician or automatic arrhythmia detection algorithms. This also corroborates earlier studies using handheld devices; these have reported AF-detection sensitivities ranging from 77% to 100% and specificities from 80% to 96% when interpreted by the physicians and sensitivities of 55%-100% and specificities of 84%-98% when the interpretation was performed using automated algorithms. 14,15,16,17,23 The abundance of health information devices and the large consumer population show an increasing interest in self-monitoring. In T A B L E 2 Atrial fibrillation (AF)-diagnosis of cardiologist-interpreted Necklace-electrocardiogram (ECG) recordings compared to two cardiologists' consensus of rhythm from Holter ECGs

| CONCLUSIONS
The Necklace-ECG recorder produces ECG signal with sufficient quality for the detection of AF with good sensitivity and specificity as evaluated both by two cardiologists and an automated arrhythmia detection algorithm. Thus, repeated ECG recording with the Necklace-ECG and an automatic arrhythmia detection algorithm might be useful in the screening of AF in a high-risk population. This wearable measuring device could provide a new and easy method for screening, identifying and diagnosing AF.