News

Article

Machine Learning Algorithm Shows Promise in Forecasting Sleep Apnea Events

Author(s):

The study presents a machine learning algorithm utilizing ECG data during CPAP titration that can effectively forecast sleep apnea events.

Duy Linh Thanh Tran, PhD Candidate

Credit: LinkedIn

Duy Linh Thanh Tran, PhD Candidate

Credit: LinkedIn

Obstructive sleep apnea (OSA) poses a significant burden on both patients and the healthcare system, with adherence to Continuous Positive Airway Pressure (CPAP) treatment often being inadequate. However, a new study offered a promising solution by using machine learning algorithms to detect sleep apnea events in advance and adjust CPAP pressure accordingly, potentially improving long-term treatment outcomes.

The study sought to preprocess 30-second ECG segments, transforming them into spectrograms using continuous wavelet transform, and generating features using the bag-of-features technique.

Specific frequency bands of 0.5–50 Hz, 0.8–10 Hz, and 8–50 Hz were also extracted to identify the most detected band.

Duy Linh Thanh Tran, International PhD Program of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan, and investigators aimed to develop a machine-learning algorithm using retrospective electrocardiogram (ECG) data and CPAP titration to forecast sleep apnea events before they occur.

A total of 4 algorithms were included: vector machine (SVM), k-nearest neighbor (KNN), decision tree (DT), and linear discriminative analysis (LDA). The systems were employed to detect sleep apnea events 30–90 seconds in advance.

The study involved preprocessing 30-second ECG segments, transforming them into spectrograms using continuous wavelet transform, and generating features using the bag-of-features technique. Specific frequency bands of 0.5–50 Hz, 0.8–10 Hz, and 8–50 Hz were also extracted to identify the most detected band.

Results of the study revealed the support vector machine (SVM) outperformed the other algorithms (KNN, LDA, and DT) across various frequency bands and leading time segments. Notably, the 8–50 Hz frequency band demonstrated the best accuracy, reaching an impressive 98.2%, with an F1-score of 0.93.

Investigators stated by effectively forecasting sleep apnea events in advance using only a single-lead ECG signal during CPAP titration, the study showcases a novel and promising approach to managing obstructive sleep apnea at home. Detecting sleep apnea events before they happen could significantly enhance adherence to CPAP treatment, leading to better long-term patient outcomes.

"Overall, our findings suggested that ECG signals at CPAP titration can be used to input into pre-OSA detection models," they wrote. "Detection models under the impact of CPAP may be closer to the real-life condition when patients use the PAP machine at home."

Detecting sleep apnea events in advance using a single-lead ECG signal during CPAP titration shows potential for improving the long-term use of CPAP treatment and, in turn, alleviating the heavy health-related burden of OSA on patients and the healthcare system, the team stated. As this approach continues to be refined and validated, it could lead to more effective home-based management of sleep apnea, ultimately enhancing the quality of life for patients suffering from this condition.

References:

  1. Linh, T. T. D., Trang, N. T. H., Lin, S.-Y., Wu, D., Liu, W.-T., & Hu, C.-J. (2023). Detection of preceding sleep apnea using ECG spectrogram during CPAP titration night: A novel machine-learning and bag-of-features framework. Journal of Sleep Research, 1– 12. https://doi.org/10.1111/jsr.13991
Related Videos
Andreas Kremer, MD, PhD, MHBA | Credit: AASLD
Andreas Kremer, MD, PhD, MHBA | Credit: AASLD
Andreas Kremer, MD, PhD, MHBA | Credit: AASLD
Andreas Kremer, MD, PhD, MHBA | Credit: AASLD
Jörn Schattenberg, MD | Credit: Novo Nordisk
Diabetes Dialogue: Diabetes Tech Updates from November 2024 | Image Credit: HCPLive
© 2024 MJH Life Sciences

All rights reserved.