31 October 2022 in Heart Rhythm O2
Multiple studies have reported on classification of raw electrocardiograms (ECGs) using convolutional neural networks (CNNs). We investigated an application-specific CNN using a custom ensemble of features designed based on characteristics of the ECG during atrial fibrillation (AF) to reduce inappropriate AF detections in implantable cardiac monitors (ICMs). An ensemble of features was developed and combined to form an input signal for the CNN. The features were based on the morphological characteristics of AF, incoherence of RR intervals, and the fact that AF begets more AF. A custom CNN model and the RESNET18 model were trained using ICM-detected AF episodes that were adjudicated to be true AF or false detections. The trained models were evaluated using a test dataset from independent patients. The training and validation datasets consisted of 31,757 AF episodes (2516 patients) and 28,506 false episodes (2126 patients). The validation set (20% randomly chosen episodes of each type) had an area under the curve of 0.996 for custom CNN (0.993 for RESNET18). Thresholds were chosen to obtain a relative sensitivity and specificity of 99.2% and 92.8%, respectively (99.2% and 87.9% for RESNET18, respectively). The performance in the independent test set (4546 AF episodes from 418 patients; 5384 false episodes from 605 patients) showed an area under the curve of 0.993 (0.991 for RESNET18) and relative sensitivity and specificity of 98.7% and 91.4%, respectively, at chosen thresholds (98.9% and 88.2% for RESNET18, respectively). An ensemble of features-based CNNs was developed that reduced inappropriate AF detection in ICMs by over 90% while preserving sensitivity.
1 August 2024 in Circulation: Arrhythmia and Electrophysiology
BACKGROUND: Atrial fibrillation (AF) events in cardiac implantable electronic devices (CIEDs) are temporally associated with stroke risk. This study explores temporal differences in AF burden associated with HF hospitalization risk in patients with CIEDs. METHODS: Patients with HF events from the Optum de-identified Electronic Health Records from 2007 to 2021 and 120 days of preceding CIED-derived rhythm data from a linked manufacturer’s data warehouse were included. AF burden ≥5.5 h/d was defined as an AF event. The AF event burden in the case period (days 1–30 immediately before the HF event) was considered temporally associated with the HF event and compared with the AF event burden in a temporally dissociated control period (days 91–120 before the HF event). The odds ratio for temporally associated HF events and the odds ratio associated with poorly rate-controlled AF (>110 bpm) were calculated. RESULTS: In total, 7257 HF events with prerequisite CIED data were included; 957 (13.2%) patients had AF events recorded only in either their case (763 [10.5%]) or control (194 [2.7%]) periods, but not both. The odds ratio for a temporally associated HF event was 3.93 (95% CI, 3.36–4.60). This was greater for an HF event with a longer stay of >3 days (odds ratio, 4.51 [95% CI, 3.57–5.68]). In patients with AF during both the control and case periods, poor AF rate control during the case period also increased HF event risk (1.78 [95% CI, 1.22–2.61]). In all, 222 of 4759 (5%) patients without AF events before their HF event had an AF event in the 10 days following. CONCLUSIONS: In a large real-world population of patients with CIED devices, AF burden was associated with HF hospitalization risk in the subsequent 30 days. The risk is increased with AF and an uncontrolled ventricular rate. Our findings support AF monitoring in CIED algorithms to prevent HF admissions.