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

OBJECTIVE: To evaluate the diagnostic performance of a deep learning system for automated detection of atrial fibrillation (AF) in photoplethysmographic (PPG) pulse waveforms. METHODS: We trained a deep convolutional neural network (DCNN) to detect AF in 17 s PPG waveforms using a training data set of 149 048 PPG waveforms constructed from several publicly available PPG databases. The DCNN was validated using an independent test data set of 3039 smartphone-acquired PPG waveforms from adults at high risk of AF at a general outpatient clinic against ECG tracings reviewed by two cardiologists. Six established AF detectors based on handcrafted features were evaluated on the same test data set for performance comparison. RESULTS: In the validation data set (3039 PPG waveforms) consisting of three sequential PPG waveforms from 1013 participants (mean (SD) age, 68.4 (12.2) years; 46.8% men), the prevalence of AF was 2.8%. The area under the receiver operating characteristic curve (AUC) of the DCNN for AF detection was 0.997 (95% CI 0.996 to 0.999) and was significantly higher than all the other AF detectors (AUC range: 0.924-0.985). The sensitivity of the DCNN was 95.2% (95% CI 88.3% to 98.7%), specificity was 99.0% (95% CI 98.6% to 99.3%), positive predictive value (PPV) was 72.7% (95% CI 65.1% to 79.3%) and negative predictive value (NPV) was 99.9% (95% CI 99.7% to 100%) using a single 17 s PPG waveform. Using the three sequential PPG waveforms in combination (<1 min in total), the sensitivity was 100.0% (95% CI 87.7% to 100%), specificity was 99.6% (95% CI 99.0% to 99.9%), PPV was 87.5% (95% CI 72.5% to 94.9%) and NPV was 100% (95% CI 99.4% to 100%). CONCLUSIONS: In this evaluation of PPG waveforms from adults screened for AF in a real-world primary care setting, the DCNN had high sensitivity, specificity, PPV and NPV for detecting AF, outperforming other state-of-the-art methods based on handcrafted features.

Diagnostic performance of an automatic blood pressure measurement device, Microlife WatchBP Home A, for atrial fibrillation screening in a real-world primary care setting.

OBJECTIVE: To evaluate the diagnostic performance of a UK National Institute for Health and Care Excellence-recommended automatic oscillometric blood pressure (BP) measurement device incorporated with an atrial fibrillation (AF) detection algorithm (Microlife WatchBP Home A) for real-world AF screening in a primary healthcare setting. SETTING: Primary healthcare setting in Hong Kong. INTERVENTIONS: This was a prospective AF screening study carried out between 1 September 2014 and 14 January 2015. The Microlife device was evaluated for AF detection and compared with a reference standard of lead-I ECG. PRIMARY OUTCOME MEASURES: Diagnostic performance of Microlife for AF detection. RESULTS: 5969 patients (mean age: 67.2±11.0 years; 53.9% female) were recruited. The mean CHA2DS2-VASc ( C : congestive heart failure [1 point]; H : hypertension [1 point]; A2 : age 65-74 years [1 point] and age ≥75 years [2 points]; D : diabetes mellitus [1 point]; S : prior stroke or transient ischemic attack [2 points]; VA : vascular disease [1 point]; and Sc : sex category [female] [1 point])score was 2.8±1.3. AF was diagnosed in 72 patients (1.21%) and confirmed by a 12-lead ECG. The Microlife device correctly identified AF in 58 patients and produced 79 false-positives. The corresponding sensitivity and specificity for AF detection were 80.6% (95% CI 69.5 to 88.9) and 98.7% (95% CI 98.3 to 98.9), respectively. Among patients with a false-positive by the Microlife device, 30.4% had sinus rhythm, 35.4% had sinus arrhythmia and 29.1% exhibited premature atrial complexes. With the low prevalence of AF in this population, the positive and negative predictive values of Microlife device for AF detection were 42.4% (95% CI 34.0 to 51.2) and 99.8% (95% CI 99.6 to 99.9), respectively. The overall diagnostic performance of Microlife device to detect AF as determined by area under the curves was 0.90 (95% CI 0.89 to 0.90). CONCLUSIONS: In the primary care setting, Microlife WatchBP Home was an effective means to screen for AF, with a reasonable sensitivity of 80.6% and a high negative predictive value of 99.8%, in addition to its routine function of BP measurement. In a younger patient population aged <65 years with a lower prevalence of AF, Microlife WatchBP Home A demonstrated a similar diagnostic accuracy.

Diagnostic Performance of a Smartphone-Based Photoplethysmographic Application for Atrial Fibrillation Screening in a Primary Care Setting.

BACKGROUND: Diagnosing atrial fibrillation (AF) before ischemic stroke occurs is a priority for stroke prevention in AF. Smartphone camera-based photoplethysmographic (PPG) pulse waveform measurement discriminates between different heart rhythms, but its ability to diagnose AF in real-world situations has not been adequately investigated. We sought to assess the diagnostic performance of a standalone smartphone PPG application, Cardiio Rhythm, for AF screening in primary care setting. METHODS AND RESULTS: Patients with hypertension, with diabetes mellitus, and/or aged ≥65 years were recruited. A single-lead ECG was recorded by using the AliveCor heart monitor with tracings reviewed subsequently by 2 cardiologists to provide the reference standard. PPG measurements were performed by using the Cardiio Rhythm smartphone application. AF was diagnosed in 28 (2.76%) of 1013 participants. The diagnostic sensitivity of the Cardiio Rhythm for AF detection was 92.9% (95% CI] 77-99%) and was higher than that of the AliveCor automated algorithm (71.4% [95% CI 51-87%]). The specificities of Cardiio Rhythm and the AliveCor automated algorithm were comparable (97.7% [95% CI: 97-99%] versus 99.4% [95% CI 99-100%]). The positive predictive value of the Cardiio Rhythm was lower than that of the AliveCor automated algorithm (53.1% [95% CI 38-67%] versus 76.9% [95% CI 56-91%]); both had a very high negative predictive value (99.8% [95% CI 99-100%] versus 99.2% [95% CI 98-100%]). CONCLUSIONS: The Cardiio Rhythm smartphone PPG application provides an accurate and reliable means to detect AF in patients at risk of developing AF and has the potential to enable population-based screening for AF.