A Comprehensive Review of ECG-Based Sleep Apnea Detection

Abstract

Sleep apnea (SA) is a typical rest problem which could disable the mortal physiological frame. Hence, early determination of SA is of extraordinary interest. The conventional fashion for diagnosing SA is a short- term polysomnography (PSG) assessment. At the point when PSG has confined availability, programmed SA webbing with a lower number of signs ought to be allowed of. The main part of this study is to produce and assess a SA position model in view of electrocardiogram (ECG) and blood oxygen absorption (SpO2). We embraced a multimodal way to deal with immingle ECG and SpO2 signals at the include position. also, highlight determination was led exercising the recursive element disposal with cross-approval (RFECV) computation and random forest (RF) classifier used to insulate between apnea and typical occasions. Tests were led on the Apnea- ECG information base. The presented computation got a perfection of 97.5, a responsiveness of 95.9, an explicitness of 98.4 and an AUC of in per-section grouping, and beat once workshop. The issues showed that ECG also, SpO2 are complementary in feting SA, and that the mix of ECG and SpO2 improves the capacity to dissect SA. Accordingly, the proposed fashion can conceivably be an option to traditional discovery strategies.

Introduction

I. INTRODUCTION

Sleep apnea (SA) is a common sleep complaint, also generally known as obstructive sleep apnea (OSA). The clinical presentation of SA is a conclusion of nasal airflow or a drop in airflow intensity by further than 30 compared to the base position, but the corresponding breathing movements are maintained. At the same time, oxygen desaturation decreases by further than 4 for further than 10 s. The prevalence of OSA in adults ranges from 9% to 38%, increasing with age Low-quality sleep accompanied by apnea generally leads directly to poor attention, memory loss, slow response, and depression. In addition, OSA is a implicit trouble to numerous physiological systems of  the mortal body, especially the cardiovascular system. It can induce hypertension, heart failure, coronary roadway complaint, diabetes, and other conditions, which seriously hang the health of cases. If patients are diagnosed and treated at an early stage of OSA, the health risks can be mitigated. Thus, timely diagnosis of cases with OSA is essential. Clinically, polysomnography (PSG) is the reference standard for the diagnosis of SA.PSG is effective in covering sleep conditions by collecting colorful physiological signals similar as electrocardiogram (ECG), electroencephalogram (EEG), electromyogram (EMG), blood oxygen desaturation (SpO2), airflow signals, respiratory trouble still, wearing too numerous detectors during physiological signal collection can beget discomfort to the case. In addition, the diagnosis of OSA requires sleep specialists to spend a lot of time manually assaying PSG data. Thus, automatic discovery of SA using smaller signals is necessary.

Experimenters have generally developed SA discovery algorithms using ECG signals. ECG is a non-invasive fashion for recording the electrical exertion of heart and the physiological exertion of heart is regulated under the autonomic nervous system (ANS). Clinically, heart rate variability (HRV) is an important index of the outgrowth of ANS regulation. Thus, it's doable to screen for apnea by covering ECG during sleep. Yet, ECG signals are fluently told by cardiovascular complaint status. This makes the diagnosis of SA more grueling. piecemeal from ECG signals, SpO2 signals are also extensively used to descry SA as the lack of respiratory effort due to SA events can lead to a drop in SpO2. repetitious oxygen desaturation is largely specific for apnea. still, the perceptivity of oximetry is generally low, as not all apnea events lead to perceptible desaturations. Therefore, SpO2 alone or ECG alone can be used as an implicit individual means of SA, but not as a dependable means. With technological advances in detectors and low-power bedded systems, the collection of physiological signals has come easier and further provident. Thus, we consider using multiple signals to develop a further dependable discovery algorithm of SA, rather than being limited to a single signal.

This study explores the effectiveness and trust ability of a multimodal approach to the automated discovery of SA events using a combined channel of ECG and SpO2.

To this end, we uprooted features from ECG signal and SpO2 signal independently, and also fused the features of the two different modalities. point selection was performed using the recursive point elimination with cross-validation (RFECV) algorithm. The named features were fed to the RF classifier to identify sleep apnea events. Our study provides three main benefactions to exploration. First, we corroborate the complementarity of ECG and SpO2 signals to automatically descry SA. When the two signals are combined, the individual capability is increased. Second, the RFECV algorithm is employed to elect the most important features. The proposed SA discovery fashion uses a lower   number of features and is computationally   affordable compared to utmost of the being styles. Third, we enrich the system in the field of the                     automated discovery  of              SA by applying  a                                 multimodal approach to fuse ECG and SpO2 signals at the point position. So far, utmost of the extant literature primarily used SpO2 alone or ECG alone, but didn't consider the combination of ECG and SpO2.

II. RELATED WORKS

Optimization of Sleep Apnea Detection using SpO2 and ANN - Repetitious respiratory disturbance during sleep is called Sleep Apnea Hypopnea Syndrome and causes colorful conditions. Different features and classifiers have been used by different experimenters to descry sleep apnea. This study is accepted to identify the better performing blood oxygen desaturation features subset using an Artificial Neural Network classifier for sleep Apnea discovery. A database of 8 subjects with one- nanosecond reflection is used to test the proposed system. The optimized system has seven features chosen from a total set of sixty- one features presenting a high delicacy rate using a inheritable algorithm. The Artificial Neural Network achieved a 97.7% accuracy rate using only seven features selected by the genetic algorithm. Another popular signal is blood oxygen desaturation (SpO2), measured by palpitation oximetry which has some added advantage over other detectors. It's further movable and can be used in hypopneas where the drop in oxygen desaturation caused by a reduction in the airflow maturity of the cases no clear ECG changes can be detected. Time sphere and time frequency sphere features of SpO2 are used with a different classifier to descry Apneas event duly by different experimenters. It's essential to find a dominant and optimum subset of features for bracket because learning with all available features could have a negative effect on performance, particularly when inapplicable or spare features are present. That's why a system optimization of the features for Artificial Neural Network (ANN) is demanded. In this work, a successful optimization of sleep Apnea discovery is carried out. It's showed that the delicacy achieves by combining the ANN with GA is relatively good. The features named by the GA algorithm are substantially time-frequency signal (four among seven) only two time and one frequency point is named. From this selection procedure, it's accessible that Apnea events have utmost of the information in time- frequency space. Compared to former work it improves the delicacy. In future, this work can be tested on different databases and classifiers. A limitation of the styles used at that time that the structure of the named MLP is calculated using a                   rule of thumb and experimental hunt which can be replaced, for illustration, by a GA algorithm and the trial isn't subject independent.

An Efficient Method to Detect Sleep Hypopnea-Apnea Events Based on EEG Signals     -          Sleep disturbances can lead to inadequate sleep at night and internal fatigue during the day, which can have a significant negative impact on our lives. The two main grueling problems in sleep analysis are sleep stage scoring and apnea- hypopnea discovery. Hypopnea refers to a 30 drop in oronasal airflow and a 4 drop in blood oxygen desaturation for further than 10 seconds, or 50 drop in nose and mouth airflow and 3 drop in blood oxygen desaturation for further than 10s. In this paper, the three types of apneas are allocated according to a reasonable proportion proposed by the medical experts of respiratory conditions from Tianjin casket Sanitarium, the OSA is 60, and the MSA and the CSA are 30 and 10, independently. Since the characteristics of hypopnea- apnea pattern are different from other conditions, the symptoms first appear during darkness sleep, and also promote the development of cardiopulmonary failure. thus, it's particularly important to descry the presence of hypopnea- apnea pattern in time. The proposed system finds the approximate entropy of the detail measure as characteristics of EEG signals during the apnea- hypopnea events. also a machine learning bracket model is established and results are attained on the test set as the final evaluation indicator of the point. Our model achieves the stylish result with delicacy94.33, perceptivity 93.10, and particularity 95.07 and provides a discovery system that uses the EEG signals to automatically descry the hypopnea- apnea events that do during darkness sleep.

Sleep Apnea Detection Using a Feed- Forward Neural Network on ECG Signal - Sleep Apnea is a sleep complaint characterized by conclusion of breathing during sleep that can last seconds or twinkles. In addition, sleep apnea is considered an important factor for morbidity and mortality due to it's direct effect on the cardiovascular system. These goods are associated with physiological functions similar as systemic hypertension and increased sympathetic exertion that compromise the heart. There are three different types of apnea central (CSA), obstructive (OSA) and mixed. This bracket is grounded on the respiratory trouble. CSA, causes the loss of all respiratory trouble during sleep and is also generally marked by diminishments in blood oxygen desaturation, while OSA is characterized by intermittent pauses in breathing during sleep caused by the inhibition of the upper airway.

The combination of these symptoms takes to mixed apnea. The symptoms associated to sleep apnea are fatigue, day somnolence, low response and visual problems. Entire population anyhow of gender, age or race can be affected by sleep apnea This paper presents a suitable and effective perpetration for detecting nanosecond grounded analysis of sleep apnea by Electrocardiogram (ECG) signal processing. Using the PhysioNet apnea- ECG database, a median sludge was applied to the recordings in order to gain the Heart Rate Variability (HRV) and the ECG- deduced respiration (EDR). The posterior uprooted features were used for training, testing and confirmation of a Artificial Neural Network (ANN). Training and testing sets were attained by aimlessly divide the data until it reaches a good performance using ak-fold cross confirmation (k = 10). According to results, the ANN bracket has sufficient delicacy for sleep apnea discovery and diagnosis (82,120). This promising early-stage result may leads to reciprocal studies including indispensable features selection styles and/ or other bracket models.

III. OBJECTIVES

1. The paper thoroughly examines the various methods and techniques used in sleep apnea detection through ECG data, providing readers with a clear understanding of the existing landscape, the evolution of techniques, and the various approaches explored in this context.
2. The paper focuses on the Random Forest algorithm, delving into its application, strengths, limitations, and instances where it has proven effective in the context of sleep apnea detection.
3. Different data sources for ECG data used in sleep apnea detection, such as in-lab polysomnography, home-based monitoring, and wearable devices, are explored, highlighting their differences and their implications on the detection process.
4. The paper investigates feature extraction methods, making it easier for readers to grasp the technical intricacies involved in translating raw ECG data into meaningful information for sleep apnea detection. Preprocessing steps for ECG data, including noise reduction, artifact removal, and data augmentation, are evaluated, helping readers understand the significance of these steps and their impact on the accuracy and reliability of detection models. Specific performance metrics, such as sensitivity, specificity, accuracy, and F1-score, are explained, allowing readers to gauge the efficacy of sleep apnea detection models and their suitability in the context of ECG-based detection.
5. The paper addresses challenges and limitations in using ECG data for sleep apnea detection, including considerations like data quality, class imbalance, and the generalizability of models in diverse scenarios.
6. A comparative analysis is provided, comparing the Random Forest algorithm with alternative machine learning and deep learning approaches used in sleep apnea detection, helping readers understand the relative advantages and disadvantages of each approach.
7. The clinical relevance of ECG-based sleep apnea detection is explored, shedding light on how this technology can impact patient diagnosis, treatment, and healthcare practices from a reader's perspective.
8. The paper  highlights emerging trends and research gaps in the field, proposing potential future directions, providing valuable insights for researchers and stakeholders looking to advance this area of study. Real world applications of ECG-based sleep apnea detection, such as telemedicine, remote patient monitoring, and personalized treatment strategies, are discussed, illustrating the practical significance of this research to readers.
9. Ethical concerns surrounding data security, patient consent, and privacy in the context of ECG-based sleep apnea detection are addressed, ensuring a well-rounded exploration of the subject matter from the reader's viewpoint.

IV. METHODOLOGY

A. Proposed Framework

1. The proposed framework involves the extraction of     the features from the two modalities – ECG signal and SpO2 signal. The features are then fused and the essential features are selected for classifying the results. The flow diagram framework is as shown in figure 1.      

B. Preprocessing

1. For the noise in the ECG signal similar as baseline drift and power frequency interference, we use FIR bandpass filter with passband of 350 Hz to denoise the original ECG signal. Also, the entire ECG signal is segmented into 1- min parts by pertaining to the reflections in the database.
2. Grounded on the per- nanosecond ECG member, we use the Hamilton algorithm to detect the R peaks, and correct the position of the R peaks to the maximum value, so as to insure the delicacy of the R peaks discovery.
3. The                                 RR interval signal is attained by the interval between consecutive R peaks, and the RR interval outliers are removed with reference to the system of    the RAMP(RAMP signal is a type of standard signal which starts at ???? = 0 and increases linearly with time signal is attained by the breadth of R surge.
4. In particular, one of the simplest approaches to gain an EDR (ECG- deduced respiration) signal is by working in the RAMP signal, so the RAMP signal is also called the EDR signal.
5. SpO2 and ECG recordings are collected contemporaneously. also, the entire SpO2 signal is resolved into 1- min parts, and parts that violated its physiological significance (SpO2 values lower than 50) are removed (25). also, the RR interval signal, RAMP signal and SpO2 signal are used for posterior point birth.

References

[1] SHEIKH SHANAWAZ MOSTAFA; JOAO PAULO CARVALHO; FERNANDO MORGADO-DIAS; ANTONIO RAVELO-GARCÍA, “Optimization of sleep apnea detection using SpO2 and ANN”, 2017 XXVI International Conference on Information, Communication and Automation Technologies (ICAT) Sarajevo, Bosnia and Herzegovina. [2] YAO WANG 1,3, SIYU JI 2 , TIANSHUN YANG 2 , XIAOHONG WANG 1 , HUIQUAN WANG 1,3, AND XIAOYUN ZHAO 4 ,“An Efficient Method to Detect Sleep Hypopnea- Apnea Events Based on EEG Signals” , 1School of Life Sciences, Tiangong University, Tianjin 300387, China 2School of Electronics and Information Engineering, Tiangong University, Tianjin 300387, China 3School of Precision Instruments and Optoelectronics Engineering, Tianjin University, Tianjin 300072, China 4Tianjin Chest Hospital, Tianjin 300000, China. [3] TANMOY PAUL1,2, OMIYA HASSAN1, KHUDER ALABOUD2,3, HUMAYERA ISLAM2,3, MD KAMRUZ ZAMAN RANA2,3, SYED K. ISLAM, PHD1, ABU S. M. MOSA, PHD1,2,3,4*, “ECG and SpO2 Signal-Based Real-Time Sleep Apnea Detection Using Feed-Forward Artificial Neural Network ”, 1Department of Electrical Engineering and Computer Science; 2 NextGen Biomedical Informatics Center; 3 Institute of Data Science and Informatics; 4 Department of Health Management and Informatics, University of Missouri, Columbia, MO, USA. [4] UDDIN MB, CHOW CM, SU SW, “Classification methods to detect sleep apnea in adults based on respiratory and oximetry signals: A systematic review”, Physiol Meas. 2018 [5] MUZA RT “Central sleep apnoea - A clinical review”, Journal of Thoracic Disease. 2015. [6] VIMALA V, RAMAR K, ETTAPPAN M, “An Intelligent Sleep Apnea Classification System Based on EEG Signals”, J Med Syst. 2019. [7] MOSTAFA SS, MENDONÇA F, MORGADO-DIAS F, RAVELO-GARCÍA A, “SpO2 based sleep apnea detection using deep learning”, In: INES 2017 - IEEE 21st International Conference on Intelligent Engineering Systems, Proceedings. 2017. [8] HAFEZI M, MONTAZERI N, SAHA S, ZHU K, GAVRILOVIC B, YADOLLAHI A, ET AL, “Sleep Apnea Severity Estimation from Tracheal Movements Using a Deep Learning Model”, IEEE Access. 2020. [9] ERDENEBAYAR U, KIM YJ, PARK JU, JOO EY, LEE KJ, “Deep learning approaches for automatic detection of sleep apnea events from an electrocardiogram”, Comput Methods Programs Biomed. 2019. [10] MOSTAFA SS, MENDONÇA F, RAVELO-GARCÍA AG, MORGADO-DIAS F, “A systematic review of detecting sleep apnea using deep learning”, Sensors (Switzerland). 2019. [11] MOSTAFA SS, MENDONÇA F, JULIÁ-SERDÁ G, MORGADO-DIAS F, RAVELO-GARCÍA AG, “SC3: self-configuring classifier combination for obstructive sleep apnea”, Neural Comput Appl. 2020. [12] JIANMING ZHU 1, AOJIE ZHOU 1, QIONG GONG 1, YU ZHOU 1 , JUNXIANG HUANG 1 AND ZHENCHENG CHEN 2,*, “Detection of Sleep Apnea from Electrocardiogram and Pulse Oximetry Signals Using Random Forest”, 1 School of Life and Environmental Sciences, Guilin University of Electronic Technology, Guilin 541004, China; 2 School of Electronic Engineering and Automation, Guilin University of Electronic Technology, Guilin 541004, China.

Copyright

Copyright © 2023 Mr. Krishna Gudi, Narasimha N, Pradhyumna K, Saahil Dharmaji, Trupthi G B. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.