This document discusses sleep apnea detection from single-lead electrocardiogram (ECG) signals using machine learning and deep learning algorithms. It analyzes 70 ECG recordings from a publicly available dataset. The document outlines existing sleep apnea detection methods, the advantages of using ECG signals, and proposes a system to collect ECG data, extract features, build models using machine and deep learning, and detect sleep apnea. Key steps include pre-processing ECG signals, extracting time, frequency and nonlinear heart rate variability features, selecting important features, and comparing different machine and deep learning models for sleep apnea classification.

1. SLEEPAPNEA DETECTION FROM SINGLE-LEAD ECG:COMPREHENSIVE
ANALYSIS OF MACHINE LEARNING AND DEEP LEARNING ALGORITHMS
Under the guidance of: presented by :
Mrs. CH. SWARNALATHA M. RADHIKA (4511-19-733-014)
K. HARI HARA VARA PRASAD (4511-19-733-032)
D. MADHU (4511-19-733-040)

2. CONTENTS
■ ABSTRACT
■ INTRODUCTION
■ EXISTING SYSTEM & DISADVANTAGES
■ PROPOSED SYSTEM & ADVANTAGES
■ SYSTEM REQUIREMENTS
■ SYSTEM ARCHITECTURE
■ MODULES
■ CONCLUSION
■ REFERENCES

3. ABSTRACT
Sleep apnea is a common sleep breathing disorder in which patients suffer from stopping or decreasing
airflow to the lungs for more than 10 seconds. Accurate detection of sleep apnea episodes is an important
step in devising appropriate therapies and management strategies. Provides a comprehensive analysis of
machine learning and deep learning algorithms on 70 recordings of the Physio Net ECG Sleep Apnea
v1.0.0 dataset. Firstly, electrocardiogram (ECG) signals were pre-processed and segmented and then
machine learning and deep learning methods were applied for sleep apnea detection. To handle our bio
signal processing requirement, all networks were similarly changed.

4. INTRODUCTION
■ Sleep accounts for about one third of the human life span and plays an important role in health and
quality of life. Two main stages of sleep are called rapid eye movement (REM)and non-rapid eye
movement (NREM).
■ During NREM, oxygen consumption and heart rate decrease as well as the blood pressure, while in the
REM stage, blood pressure and heart rate surge.
■ Sleep apnea may occur during any stage of sleep, but it is most dominant during the REM. This is
mainly because the muscles in the upper airway are further relaxed during the REM stage of sleep.
■ The main aim to facilitate researchers with different ML and DL techniques for sleep apnea detection.
To classify and characterize sleep apnea ECG signals are applied as input.

5. EXISTING SYSTEM
■ Image processing techniques, and signal analysis methods are the main approaches for sleep
apnea monitoring. Medical image analysis is a capable method for analyzing sleep apnea
especially in severe sleep apnea. Medical image analysis also provides a unique opportunity for
screening anatomical changes during sleep apnea.
■ Polysomnography (PSG) is a gold standard for sleep apnea monitoring. PSG consists of various
biological signals such as electrocardiogram, electroencephalogram, airflow pulse oximetry,
arterial blood oxygen saturation, and nasal flow.

6. DISADVANTAGES
■ However, none of these techniques can be performed in non-specialist settings such as in home and
conducting PSG is expensive and often unavailable due to the shortage of physical therapists and sleep
monitoring units.
■ Polysomnography (PSG). However, this method requires many electrodes and wires, as well as an
expert to monitor the test.

7. PROPOSED SYSTEM
■ We have proposed instead using a single channel signal for SA diagnosis. Among these options,
the ECG signal is one of the most physiologically relevant signals of SA occurrence
■ ECG signal-based methods mainly use features (i.e. frequency domain, time domain, and other
nonlinear features) acquired from ECG and its derived signals in order to construct the model
■ Conventional machine learning and novel deep learning (convolutional network and recurrent
networks and hybrid convolutional-recurrent networks) are two main automatic approaches for
sleep apnea monitoring from physiological signals.

8. ADVANTAGES
■ ECG can be easily recorded, ECG-based portable devices represent a better option. which are
simple, of low-cost.
■ Using ECG signals can greatly reduce the complexity of diagnostic SA tests and allow for better
monitoring of physiological changes in the patient.
■ Hybrid deep models, which have the best accuracy, sensitivity, and specificity, yield the best
detection performance. Deep learning methods rely on automatic feature extraction.

9. SYSTEM REQUIREMENTS
■ Hardware Requirements
■ System Processor: Core i3 or above
■ Hard Disk: 250 GB minimum
■ Ram: 4 GB or higher
■ Software Requirements
■ Operating system: Windows 10 or higher
■ Programming Language: Python
■ Framework: Anaconda
■ IDE: Jupiter Notebook

10. SYSTEM ARCHITECTURE

11. MODULES
■ DATA COLLECTION
■ DATA PRE-PROCESSING
■ FEATURE EXTRACTION
■ FEATURE SELECTION
■ MODEL BUILDING
■ MODEL EVALUATION
■ SLEAP APNEA DETECTION

12. DATA COLLECTION
 Data collection is process of collecting the data from different source. The dataset is collected
from kaggle.
 Physio Net Apnea-ECG Database v1.0.0 was used to build our models and to compare their
performances. The database contains 70 recordings from 32 individuals , divided into four
groups: A, B, C, and X.
 The length of the 70 recordings was 8.2±0.52 hours. ECG signals are digitized at 100 Hz. The
annotation of the presence of sleep apnea is provided for every 1-min by a sleep expert.

13. DATA PREPROCESSING
■ ECG signals were segmented into 1-min intervals. To extract the R-R Intervals from the ECG
signal, the Hamilton R-peak detection method based on an open-source code was used. A
median filter was used for removing physiologically uninterpretable points.
■ The extracted R-R Intervals were then presented to the developed machine learning algorithms
after appropriate processing .
■ In addition to the R-R Intervals, the amplitudes of R-peaks were also extracted and presented
to the deep models. Cubic interpolation at 3Hz was applied to sample R-R intervals and R-peak
amplitudes at an equal rate. Finally, the interpolated R-R Intervals and R-peak amplitudes were
fed to the deep models.

14. FEATURE EXTRACTION
■ Feature extraction plays a vital role in boosting the performance of proposed algorithms.
■ The most important features of ECG for apnea detection are HRV parameters .
■ In this, a comprehensive set of time-domain, frequency-domain, and non-linear features of HRV
were extracted and presented to different machine learning models for sleep apnea detection.

15. FEATURE SELECTION
■ Many irrelevant features lead to the curse of dimensionality problem and have a detrimental
impact on the performance of machine learning methods.
■ To cope with these problems, feature selection techniques are applied before classification.
■ Feature selection provides a unique opportunity for detecting the most effective and important
features for classification. Here, we applied principal component analysis (PCA) for dimension
reduction

16. MODEL BUILDING

19. CONCLUSION
We have provided a comprehensive comparison between different machine learning and deep
learning algorithms, in a unified framework, for the detection of sleep apnea from single-lead ECG.
We have completed the data collection module and extracted the Physio Net Apnea-ECG dataset
from kaggle.

20. REFERENCES
1. Physio Net Apnea -ECG: https://physionet.org/content/apnea -ecg /1.0.0/;
UCDDB: https://archive.physionet.org/physiobank/database/ucddb/.
2. https://www.kaggle.com/code/paulopinheiro/apnea-ecg/data
3. M. Bahrami and M. Forouzanfar , "Detection of sleep apnea from single-lead ECG:
Comparison of deep learning algorithms," in 2021 IEEE International Symposium on Medical
Measurements and Applications (Me Me), 2021: IEEE, pp. 1-5.
4. N. Pombo, B. M. Silva, A. M. Pinho , and N. Garcia, "Classifier precision analysis for sleep
apnea detection using ecg signals," IEEE Access, vol. 8, pp. 200477-200485, 2020.
5. M. H. Kryger , T. Roth, and W. C. Dement, Principles and Practice of sleep medicine, 6th
Edition ed. Elsevier Inc., 2017.

21. THANK YOU