# Sleep Apnea Detection Based on Multi-Scale Residual Network

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

_{2}) and other physiological signals [10,11], and then SA is manually labeled. The detection process is complex and costly, making it impossible for many patients to be diagnosed and treated in a timely manner. Therefore, it has become a consensus of researchers to explore convenient and inexpensive methods of detecting SA.

## 2. Materials and Methods

#### 2.1. Flow Diagram of the Work

#### 2.2. Experimental Data

#### 2.3. Signal Denoising

- Baseline wandering—It is mainly caused by the low-frequency interference signals caused by poor contact of the measuring electrode or the patient’s breathing [21]. The frequency is between 0.05 Hz and 2 Hz, indicating that the ECG signals deviate from the normal baseline position.
- Power line interference—It is mainly 50 Hz/60 Hz noise generated by the power system, which will cause the entire waveform to be ambiguous and have a greater impact on the waveform.
- Electromyography noise—It is mainly caused by muscle fibrillation and contraction. The amplitude is small and the frequency is high [22]. The frequency is between 5 Hz and 2000 Hz, presenting an irregular and rapidly changing waveform.

#### 2.4. R Peak Location and Signal Extraction

#### 2.5. Residual Network

#### 2.6. Construction of Multi-Scale Residual Network Model

#### 2.7. Data Imbalance Processing

## 3. Experiment and Result Analysis

#### 3.1. Sleep Apnea Detection Experiment

#### 3.2. Per-Recording Classification

#### 3.3. Test the Model on the UCD Database

#### 3.4. Comparison of Similar Research Results

## 4. Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## References

- Sharma, M.; Patel, V.; Acharya, U.R. Automated identification of insomnia using optimal bi-orthogonal wavelet transform technique with single-channel EEG signals. Knowl.-Based Syst.
**2021**, 224, 107078–107083. [Google Scholar] [CrossRef] - Sharma, M.; Patel, V.; Tiwari, J.; Acharya, U. Automated Characterization of Cyclic Alternating Pattern Using Wavelet-Based Features and Ensemble Learning Techniques with EEG Signals. Diagnostics
**2021**, 11, 1380. [Google Scholar] [CrossRef] - Aurora, R.N.; Crainiceanu, C.; Gottlieb, D.J.; Kim, J.S.; Punjabi, N.M. Obstructive Sleep Apnea during REM Sleep and Cardiovascular Disease. Am. J. Respir. Crit. Care Med.
**2018**, 197, 653–660. [Google Scholar] [CrossRef] - Butler, M.P.; Emch, J.T.; Rueschman, M.; Sands, S.A.; Shea, S.A.; Wellman, A.; Redline, S. Apnea–Hypopnea Event Duration Predicts Mortality in Men and Women in the Sleep Heart Health Study. Am. J. Respir. Crit. Care Med.
**2018**, 199, 903–912. [Google Scholar] [CrossRef] - Benjafield, A.V.; Ayas, N.T.; Eastwood, P.R.; Heinzer, R.; Ip, M.S.M.; Morrell, M.J.; Nunez, C.M.; Patel, S.R.; Penzel, T.; Pépin, J.-L.; et al. Estimation of the global prevalence and burden of obstructive sleep apnoea: A literature-based analysis. Lancet Respir. Med.
**2019**, 7, 687–698. [Google Scholar] [CrossRef] [Green Version] - Mostafa, S.S.; Morgado-Dias, F.; Ravelo-García, A.G. Comparison of SFS and mRMR for oximetry feature selection in obstructive sleep apnea detection. Neural Comput. Appl.
**2018**, 32, 15711–15731. [Google Scholar] [CrossRef] - Romero-Ortuno, R.; O’Shea, D. Fitness and frailty: Opposite ends of a challenging continuum! Will the end of age discrimination make frailty assessments an imperative? Age Ageing
**2013**, 42, 279–280. [Google Scholar] [CrossRef] [Green Version] - Jezzini, A.; Ayache, M.; Ibrahim, Z.; Elkhansa, L. ECG Classification for sleep apnea detection. In Proceedings of the Third International Conference on Advances in Biomedical Engineering (ICABME15), Beirut, Lebanon, 16–18 September 2015. [Google Scholar]
- Peppard, P.E.; Young, T.; Barnet, J.H.; Palta, M.; Hagen, E.W.; Hla, K.M. Increased Prevalence of Sleep-Disordered Breathing in Adults. Am. J. Epidemiol.
**2013**, 177, 1006–1014. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Sharma, H.; Sharma, K.K. An algorithm for sleep apnea detection from single-lead ECG using Hermite Basis functions. Comput. Biol. Med.
**2016**, 77, 116–124. [Google Scholar] [CrossRef] - Miao, K.; Ji, K.; Leng, X.; Xing, X.; Zou, H. Synthetic Aperture Radar Target Recognition with Feature Fusion Based on a Stacked Autoencoder. Sensors
**2017**, 17, 192. [Google Scholar] - Ap, A.; Np, A.; Bmcsa, C.; Kb, B.; Ng, D. Towards an accurate sleep apnea detection based on ECG signal: The quintessential of a wise feature selection. Appl. Soft Comput.
**2019**, 83, 105568. [Google Scholar] - Viswabhargav, C.; Tripathy, R.; Acharya, U.R. Automated detection of sleep apnea using sparse residual entropy features with various dictionaries extracted from heart rate and EDR signals. Comput. Biol. Med.
**2019**, 108, 20–30. [Google Scholar] [CrossRef] - Feng, K.; Qin, H.; Wu, S.; Pan, W.; Liu, G. A Sleep Apnea Detection Method Based on Unsupervised Feature Learning and Single-Lead Electrocardiogram. IEEE Trans. Instrum. Meas.
**2021**, 70, 1–12. [Google Scholar] [CrossRef] - Sharma, M.; Agarwal, S.; Acharya, U.R. Application of new class of antisymmetric wavelet filter banks for obstructive sleep apnea diagnosis using ECG signals. Comput. Biol. Med.
**2018**, 100, 100–113. [Google Scholar] [CrossRef] - Li, K.; Pan, W.; Li, Y.; Jiang, Q.; Liu, G. A method to detect sleep apnea based on deep neural network and hidden Markov model using single-lead ECG signal. Neurocomputing
**2018**, 294, 94–101. [Google Scholar] [CrossRef] - Urtnasan, E.; Park, J.-U.; Joo, E.Y.; Lee, K.-J. Automated Detection of Obstructive Sleep Apnea Events from a Single-Lead Electrocardiogram Using a Convolutional Neural Network. J. Med. Syst.
**2018**, 42, 104. [Google Scholar] [CrossRef] - Chen, L.; Zhang, X.; Song, C. An Automatic Screening Approach for Obstructive Sleep Apnea Diagnosis Based on Single-Lead Electrocardiogram. IEEE Trans. Autom. Sci. Eng.
**2015**, 12, 106–115. [Google Scholar] [CrossRef] - Sharma, M.; Raval, M.; Acharya, U.R. A new approach to identify obstructive sleep apnea using an optimal orthogonal wavelet filter bank with ECG signals. Inform. Med. Unlocked
**2019**, 16, 100170. [Google Scholar] [CrossRef] - Sharma, M.; Kumbhani, D.; Yadav, A.; Acharya, U.R. Automated Sleep apnea detection using optimal duration-frequency concentrated wavelet-based features of pulse oximetry signals. Appl. Intell.
**2021**, 1–13. [Google Scholar] [CrossRef] - Kiranyaz, S.; Ince, T.; Gabbouj, M. Real-Time Patient-Specific ECG Classification by 1-D Convolutional Neural Networks. IEEE Trans. Biomed. Eng.
**2016**, 63, 664–675. [Google Scholar] [CrossRef] - Yea-Hoon, K.; Sae-Byuk, S.; Shin-Dug, K. Electroencephalography Based Fusion Two-Dimensional (2D)-Convolution Neural Networks (CNN) Model for Emotion Recognition System. Sensors
**2018**, 18, 1383. [Google Scholar] - Pan, J.; Tompkins, W.J. A Real-Time QRS Detection Algorithm. IEEE Trans. Biomed. Eng.
**2007**, BME-32, 230–236. [Google Scholar] [CrossRef] - Hamilton, P. Open Source ECG Analysis. Comput. IEEE Cardiol.
**2002**, 29, 101–104. [Google Scholar] - Parhi, K.K.; Ayinala, M. Low-Complexity Welch Power Spectral Density Computation. IEEE Trans. Circuits Syst. I Regul. Pap.
**2014**, 61, 172–182. [Google Scholar] [CrossRef] - Neunhoeffer, M.; Sternberg, S. How Cross-Validation Can Go Wrong and What to Do About It. Political Anal.
**2019**, 27, 101–106. [Google Scholar] [CrossRef] [Green Version] - Cokelaer, T.; Hasch, J. ‘Spectrum‘: Spectral Analysis in Python. J. Open Source Softw.
**2017**, 2, 348. [Google Scholar] [CrossRef] - Song, C.; Liu, K.; Zhang, X.; Chen, L.; Xian, X. An Obstructive Sleep Apnea Detection Approach Using a Discriminative Hidden Markov Model From ECG Signals. IEEE Trans. Biomed. Eng.
**2016**, 63, 1532–1542. [Google Scholar] [CrossRef] - Zhang, Y.; Sun, Y.; Phillips, P.; Liu, G.; Zhou, X.; Wang, S. A Multilayer Perceptron Based Smart Pathological Brain Detection System by Fractional Fourier Entropy. J. Med. Syst.
**2016**, 40, 1–11. [Google Scholar] [CrossRef] - He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 27–30 June 2016. [Google Scholar]
- Lin, T.Y.; Goyal, P.; Girshick, R.; He, K.; Dollár, P. Focal loss for dense object detection. In Proceedings of the 2017 IEEE International Conference on Computer Vision, Venice, Italy, 22–29 October 2017; pp. 2999–3007. [Google Scholar]
- Martín-González, S.; Navarro-Mesa, J.L.; Juliá-Serdá, G.; Kraemer, J.F.; Wessel, N.; Ravelo-García, A.G. Heart rate variability feature selection in the presence of sleep apnea: An expert system for the characterization and detection of the disorder. Comput. Biol. Med.
**2017**, 91, 47–58. [Google Scholar] [CrossRef] - Surrel, G.; Aminifar, A.; Rincon, F.; Murali, S.; Atienza, D. Online Obstructive Sleep Apnea Detection on Medical Wearable Sensors. IEEE Trans. Biomed. Circuits Syst.
**2018**, 12, 762–773. [Google Scholar] [CrossRef] - Chang, H.-Y.; Yeh, C.-Y.; Lee, C.-T.; Lin, C.-C. A Sleep Apnea Detection System Based on a One-Dimensional Deep Convolution Neural Network Model Using Single-Lead Electrocardiogram. Sensors
**2020**, 20, 4157. [Google Scholar] [CrossRef] - Sinam, A.; Majumder, S. A novel approach osa detection using single-lead ecg scalogram based on deep neural network. J. Mech. Med. Biol.
**2019**, 19, 1950026. [Google Scholar]

**Figure 5.**Performance comparison of focus loss function, class weight and without using any data imbalance technology. (

**a**) The proposed method; (

**b**) without using any data imbalance technology method; (

**c**) class weight method.

Layer | Output Size | Network Architecture |
---|---|---|

conv1 | 100 × 1 | Convolutional layer: 7 × 1, 64, Stride: 3 |

conv2_ms | 50 × 1 | Pooling layer: 3 × 1, Stride: 2 |

${\left[\begin{array}{cc}{\left\{\begin{array}{c}3\times 1,16\\ 5\times 1,16\\ 7\times 1,16\\ 9\times 1,16\end{array}\right\}}_{C}& {\left\{\begin{array}{c}3\times 1,16\\ 5\times 1,16\\ 7\times 1,16\\ 9\times 1,16\end{array}\right\}}_{C}\end{array}\right]}_{T}\times 2$ | ||

conv3_ms | 25 × 1 | ${\left[\begin{array}{cc}{\left\{\begin{array}{c}3\times 1,32\\ 5\times 1,32\\ 7\times 1,32\\ 9\times 1,32\end{array}\right\}}_{C}& {\left\{\begin{array}{c}3\times 1,32\\ 5\times 1,32\\ 7\times 1,32\\ 9\times 1,32\end{array}\right\}}_{C}\end{array}\right]}_{T}\times 2$ |

conv4_ms | 13 × 1 | ${\left[\begin{array}{cc}{\left\{\begin{array}{c}3\times 1,64\\ 5\times 1,64\\ 7\times 1,64\\ 9\times 1,64\end{array}\right\}}_{C}& {\left\{\begin{array}{c}3\times 1,64\\ 5\times 1,64\\ 7\times 1,64\\ 9\times 1,64\end{array}\right\}}_{C}\end{array}\right]}_{T}\times 2$ |

conv5_ms | 7 × 1 | ${\left[\begin{array}{cc}{\left\{\begin{array}{c}3\times 1,128\\ 5\times 1,128\\ 7\times 1,128\\ 9\times 1,128\end{array}\right\}}_{C}& {\left\{\begin{array}{c}3\times 1,128\\ 5\times 1,128\\ 7\times 1,128\\ 9\times 1,128\end{array}\right\}}_{C}\end{array}\right]}_{T}\times 2$ |

1 × 1 | Dropout: 0.5, | |

Computing power | 0.144 × 10^{9} |

Forecast Result | Accuracy/% | Sensitivity/% | Specificity/% | ||||
---|---|---|---|---|---|---|---|

N | AH | Total | |||||

Realitylabel | N | 9158 | 1353 | 10,511 | 86.0 | 84.1 | 87.1 |

AH | 1036 | 5462 | 6498 | ||||

Total | 10,194 | 6815 | 17,009 |

Method | Accuracy/% | Sensitivity/% | Specificity/% | AUC% | F1-Score/% |
---|---|---|---|---|---|

ResNet | 84.6 | 82.2 | 86.1 | 0.918 | 80.3 |

ResNet + Multiscale | 86.0 | 84.1 | 87.1 | 0.931 | 82.1 |

Method | Accuracy/% | Sensitivity/% | Specificity/% | AUC | Corr/% |
---|---|---|---|---|---|

ResNet | 91.2 | 100 | 75 | 0.985 | 0.945 |

ResNet + Multiscale | 97.1 | 100 | 91.7 | 1 | 0.956 |

Method | Accuracy/% | Sensitivity/% | Specificity/% |
---|---|---|---|

ResNet | 67.1 | 35.5 | 72.2 |

ResNet + Multiscale | 72.4 | 36.5 | 83.6 |

Work | Method | Accuracy/% | Sensitivity/% | Specificity/% |
---|---|---|---|---|

Sharma and Sharma | LS-SVM | 83.4 | 79.5 | 88.4 |

Pinho et al. | ANN/SVM | 82.1 | 88.4 | 72.3 |

Viswabhargav et al. | SVM | 78.1 | 78.0 | 78.1 |

Surrel et al. | LS-SVM | 82.2 | 73.3 | 87.6 |

Li et al. | DNN + HMM | 84.7 | 88.9 | 82.1 |

Feng et al. | TDCS | 85.1 | 86.2 | 84.4 |

Martin-Gonzalez et al. | LDA + QDA + LR | 84.8 | 81.5 | 86.8 |

Chang et al. | 1D CNN | 87.9 | 81.1 | 92.0 |

Singh et al. | CNN + Decision Fusion | 86.2 | 90.0 | 83.8 |

Our method | ResNet + Multiscale | 86.0 | 84.1 | 87.1 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Fang, H.; Lu, C.; Hong, F.; Jiang, W.; Wang, T.
Sleep Apnea Detection Based on Multi-Scale Residual Network. *Life* **2022**, *12*, 119.
https://doi.org/10.3390/life12010119

**AMA Style**

Fang H, Lu C, Hong F, Jiang W, Wang T.
Sleep Apnea Detection Based on Multi-Scale Residual Network. *Life*. 2022; 12(1):119.
https://doi.org/10.3390/life12010119

**Chicago/Turabian Style**

Fang, Hengyang, Changhua Lu, Feng Hong, Weiwei Jiang, and Tao Wang.
2022. "Sleep Apnea Detection Based on Multi-Scale Residual Network" *Life* 12, no. 1: 119.
https://doi.org/10.3390/life12010119