How Effectively Can ECG Signals Alone Be Used to Detect Sleep Apnea Events in Patients?

Abstract: 

Sleep apnea is a common disorder characterized by pauses in breathing during sleep, which can lead to serious health complications if left undiagnosed. This study investigates the effectiveness of using only ECG signals to detect sleep apnea events, aiming to evaluate an alternative, non-invasive diagnostic approach. The research utilizes the Apnea-ECG Database, which includes ECG data from 35 patients, where features such as heart rate, RR interval, and frequency domain metrics were extracted from filtered ECG signals. Supervised machine learning models, including Naive Bayes (NB), Support Vector Machines (SVM), and K-Nearest Neighbors (KNN), were employed, with hyperparameter tuning and cross-validation applied to optimize performance. The NB model achieved a weighted average F1-score of 90%, indicating its effectiveness in identifying apnea events. These findings indicate that ECG-based detection may serve as a promising tool for detecting sleep apnea, with future research focusing on expanding the dataset, integrating additional physiological features, and validating the approach in clinical environments.