December 4, 2022Open Access
Detection of Left Ventricular Systolic Dysfunction from Single-Lead Electrocardiography Adapted for Wearable Devices
Resultado clave
A noise-adapted artificial intelligence model significantly improved the detection of left ventricular systolic dysfunction from noisy single-lead ECGs compared to a standard model (AUROC 0.87 vs 0.72).
Diseño del estudio
Tipo
Observational (n=116,210)
Multicéntrico
No
PICO estructurado
Does a noise-adapted deep learning model improve the detection of left ventricular systolic dysfunction from noisy single-lead ECGs compared to a standard model?
P
Población
116,210 unique patients (contributing 385,601 ECG-TTE pairs) who underwent clinically indicated 12-lead ECG with paired transthoracic echocardiogram within 15 days at Yale-New Haven Hospital. Median age 68 years, 43.7% female.
I
Intervención
Noise-adapted convolutional neural network (CNN) model trained on single-lead ECGs (Lead I) augmented with random gaussian noise within four distinct frequency ranges (3-12 Hz, 12-50 Hz, 50-100 Hz, 100-150 Hz) to emulate real-world noise sources.
C
Comparador
Standard convolutional neural network (CNN) model trained on original (clean) single-lead ECGs without noise augmentation.
O
Resultado
Detection of left ventricular systolic dysfunction (LVSD), defined as left ventricular ejection fraction (LVEF) < 40% on paired echocardiogram.surrogate
A noise-adapted deep learning model trained on augmented single-lead ECGs maintains high accuracy for detecting left ventricular systolic dysfunction even in the presence of significant real-world noise, enabling reliable use in wearable devices.
Resultado numérico
Tasa de eventos absoluta: 0.87% vs 0.72%
valor p: p=<0.001
Limitaciones
- Developed using ECGs from patients who had both an ECG and a clinically indicated echocardiogram, which differs from the intended screening population.
- Model performance may vary by the severity of LV systolic dysfunction, particularly near the 40% LVEF cutoff.
- Cannot ascertain whether the model maintains performance on every type and magnitude of noise possible on wearable devices.
Resumen
ABSTRACT Artificial intelligence (AI) can detect left ventricular systolic dysfunction (LVSD) from electrocardiograms (ECGs). Wearable devices could allow for broad AI-based screening but frequently obtain noisy ECGs. We report a novel strategy that automates the detection of hidden cardiovascular diseases, such as LVSD, adapted for noisy single-lead ECGs obtained on wearable and portable devices. Overall, 385,601 ECGs were used for development of a standard and noise-adapted model. For the noise-adapted model, ECGs were augmented during training with random gaussian noise within four distinct frequency ranges, each emulating real-world noise sources. Both models performed comparably on clean ECGs with an AUROC of 0.90. The noise-adapted model performed significantly better on the same test set augmented with four distinct real-world noise recordings at multiple signal-to-noise ratios (SNRs), including noise isolated from a portable device ECG. The standard and noise-adapted models had an AUROC of 0.72 and 0.87, respectively when evaluated on ECGs augmented with portable ECG device noise at an SNR of 0.5. This approach represents a novel strategy for the development of wearable adapted tools from clinical ECG repositories.