Application of Machine Learning Techniques for the Diagnosis of Obstructive Sleep Apnea/Hypopnea Syndrome
Abstract
:1. Introduction
2. Materials and Methods
2.1. Dataset and Ethics Declarations
2.2. Models for the Diagnosis of OSAHS
2.2.1. Model Proposed Utilizing Neural Networks
2.2.2. Model Proposed Utilizing Decision Trees
2.2.3. Proposed Model Utilizing Random Forests
2.2.4. Proposed Model Utilizing Extra Trees
3. Results
3.1. Evaluation of Models for OSAHS Diagnosis
3.2. Web Application for OSAHS Diagnosis
4. Discussion
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Benjafield, A.; Ayas, N.; Eastwood, P.; Heinzer, R.; Ip, M.; Morrell, M.; Nunez, C.; Patel, S.; Penzel, T.; Pépin, J.; et al. Estimation of the global prevalence and burden of obstructive sleep apnea: A literature-based analysis. Lancet Respir. Med. 2019, 7, 687–698. [Google Scholar] [CrossRef] [PubMed] [PubMed Central]
- Ingram, D.; Lee-Chiong, T.; Londoño, D. Costos e impacto económico del síndrome de apnea-hipopnea obstructiva del sueño (SAHOS) para la salud pública. Rev. Fac. Med. 2017, 65 (Suppl. S1), 141–147. [Google Scholar] [CrossRef]
- Barashi, N.S.; Ruiz, R.E.; Marín, L.; Ruiz, P.; Amado, S.; Ruiz Álvaro, J.; Hidalgo, P. Síndrome de apnea/hipopnea obstructiva del sueño y su asociación con las enfermedades cardiovasculares. Rev. Colomb. Cardiol. 2015, 22, 81–87. [Google Scholar] [CrossRef]
- Hidalgo-Martínez, P.; Lobelo, R. Epidemiología mundial, latinoamericana y colombiana y mortalidad del síndrome de apnea-hipopnea obstructiva del sueño (SAHOS). Rev. Fac. Med. 2017, 65 (Suppl. S1), 17–20. [Google Scholar] [CrossRef]
- Barrera, A. Síndrome de apnea hipopnea obstructiva del sueño y consumo de benzodiacepinas en adultos mayores. Reporte de un caso. Rev. Fac. Med. 2018, 61, 21–25. [Google Scholar]
- Lloberes, P.; Durán-Cantolla, J.; Martínez-García, M.Á.; Marín, J.M.; Ferrer, A.; Corral, J.; Masa, J.F.; Parra, O.; Alonso-Álvarez, M.L.; Terán-Santos, J. Diagnóstico y tratamiento del síndrome de apneas-hipopneas del sueño. Arch. Bronconeumol. 2011, 47, 143–156. [Google Scholar] [CrossRef] [PubMed]
- Mencar, C.; Gallo, C.; Mantero, M. Application of machine learning to predict obstructive sleep apnea syndrome severity. Health Inform. J. 2020, 26, 298–317. [Google Scholar] [CrossRef] [PubMed]
- Wang, K.J.; Chen, K.H.; Huang, S.H. A Prognosis Tool Based on Fuzzy Anthropometric and Questionnaire Data for Obstructive Sleep Apnea Severity. J. Med. Syst. 2016, 40, 110. [Google Scholar] [CrossRef]
- Chawla, N.V.; Bowyer, K.W.; Hall, L.O.; Kegelmeyer, W.P. SMOTE: Synthetic minority oversampling technique. J. Artif. Intell. Res. 2002, 16, 321–357. [Google Scholar] [CrossRef]
- Karamanli, H.; Yalcinoz, T.; Yalcinoz, M.A.; Yalcinoz, T. A prediction model based on artificial neural networks for the diagnosis of obstructive sleep apnea. Sleep Breath. 2016, 20, 509–514. [Google Scholar] [CrossRef] [PubMed]
- Rodrigues, P.; Santos, D.; Leite, L. Obstructive Sleep Apnea Diagnosis: The Bayesian Network Model Revisited. In Proceedings of the 2015 IEEE 28th International Symposium on Computer-Based Medical Systems, Sao Carlos, Brazil, 22–25 June 2015; pp. 115–120. [Google Scholar] [CrossRef]
- Leite, L.; Costa-Santos, C.; Rodrigues, P. Can We Avoid Unnecessary Polysomnographies in the Diagnosis of Obstructive Sleep Apnea? A Bayesian Network Decision Support Tool. In Proceedings of the 2014 IEEE 27th International Symposium on Computer-Based Medical Systems, New York, NY, USA, 27–29 May 2014; pp. 28–33. [Google Scholar] [CrossRef]
- Musman, S.; Passos, V.M.; Silva, I.B.; Barreto, S.M. Evaluation of a prediction model for sleep apnea in patients submitted to polysomnography. J. Bras. Pneumol. 2011, 37, 75–84, (In English and Portuguese). [Google Scholar] [CrossRef] [PubMed]
- Levy, J.; Álvarez, D.; Del Campo, F.; Behar, J.A. Deep learning for obstructive sleep apnea diagnosis based on single channel oximetry. Nat. Commun. 2023, 14, 4881. [Google Scholar] [CrossRef] [PubMed]
- Hemrajani, P.; Dhaka, V.S.; Rani, G.; Shukla, P.; Bavirisetti, D.P. Efficient Deep Learning Based Hybrid Model to Detect Obstructive Sleep Apnea. Sensors 2023, 23, 4692. [Google Scholar] [CrossRef] [PubMed]
- Li, F.; Xu, Y.; Chen, J.; Lu, P.; Zhang, B.; Cong, F. A deep learning model developed for sleep apnea detection: A multi-center study. Biomed. Signal Process. Control 2023, 85, 104689. [Google Scholar] [CrossRef]
- Setiawan, F.; Lin, C.W. A Deep Learning Framework for Automatic Sleep Apnea Classification Based on Empirical Mode Decomposition Derived from Single-Lead Electrocardiogram. Life 2022, 12, 1509. [Google Scholar] [CrossRef] [PubMed]
- McCulloch, W.S.; Pitts, W. A logical calculus of the ideas immanent in nervous activity. Bull. Math. Biophys. 1943, 5, 115–133. [Google Scholar] [CrossRef]
- Wu, X.; Kumar, V.; Quinlan, J.R.; Ghosh, J.; Yang, Q.; Motoda, H.; McLachlan, G.J.; Ng, A.; Liu, B.; Yu, P.S.; et al. Top 10 algorithms in data mining. Knowl. Inf. Syst. 2007, 14, 1–37. [Google Scholar] [CrossRef]
- Ho, T.K. Random decision forests. In Proceedings of the 3rd International Conference on Document Analysis and Recognition, Montreal, QC, Canada, 14–16 August 1995; Volume 1, pp. 278–282. [Google Scholar]
- Geurts, P.; Ernst, D.; Wehenkel, L. Extremely randomized trees. Mach. Learn. 2006, 63, 3–42. [Google Scholar] [CrossRef]
- Huang, W.C.; Lee, P.L.; Liu, Y.T.; Chiang, A.A.; Lai, F. Support vector machine prediction of obstructive sleep apnea in a large-scale Chinese clinical sample. Sleep 2020, 43, zsz295. [Google Scholar] [CrossRef]
- Maniaci, A.; Riela, P.M.; Iannella, G.; Lechien, J.R.; La Mantia, I.; De Vincentiis, M.; Cammaroto, G.; Calvo-Henriquez, C.; Di Luca, M.; Chiesa Estomba, C.; et al. Machine Learning Identification of Obstructive Sleep Apnea Severity through the Patient Clinical Features: A Retrospective Study. Life 2023, 13, 702. [Google Scholar] [CrossRef]
- Han, H.; Oh, J. Application of various machine learning techniques to predict obstructive sleep apnea syndrome severity. Sci. Rep. 2023, 13, 6379. [Google Scholar] [CrossRef] [PubMed]
- Ahbab, S.; Ataoğlu, H.E.; Tuna, M.; Karasulu, L.; Cetin, F.; Temiz, L.U.; Yenigün, M. Neck circumference, metabolic syndrome and obstructive sleep apnea syndrome; evaluation of possible linkage. Med. Sci. Monit. 2013, 19, 111–117. [Google Scholar] [CrossRef] [PubMed] [PubMed Central]
Index | Variable | Definition | Data Type |
---|---|---|---|
1 | Age | Age of the patient | Positive Integer |
2 | Sex | Gender of the patient | 1—Masculine 0—Feminine |
3 | Smoking | Smoking status | Boolean |
4 | AHT | Presence or absence of diagnosed arterial hypertension | Boolean |
5 | DM | Diabetes Mellitus diagnosis status | Boolean |
6 | COPD | Chronic Obstructive Pulmonary Disease status | Boolean |
7 | Asthma | Presence or absence of diagnosed asthma | Boolean |
8 | Rhinitis | Presence or absence of diagnosed rhinitis | Boolean |
9 | GERD | Presence or absence of diagnosed Gastroesophageal Reflux Disease | Boolean |
10 | Deviated septum | Presence or absence of nasal septum deviation diagnosis | Boolean |
11 | CHF | Presence or absence of diagnosed chronic heart failure | Boolean |
12 | Coronary artery disease | Presence or absence of diagnosed coronary artery disease | Boolean |
13 | CVA | Ever diagnosed or not with stroke | Boolean |
14 | Arrhythmias | Presence or absence of diagnosed arrhythmias | Boolean |
15 | Epworth | Epworth Sleepiness Scale Score | Positive Integer |
16 | Mallampati | Mallampati Score rating | Real number |
17 | Tonsils | Brodsky Classification | Positive Integer |
18 | BMI | Body Mass Index | Real Number |
19 | OSAHS | OSAHS diagnosis | 0—OSAHS negative 1—OSAHS positive |
Characteristics | n = 601 (%) |
---|---|
Age—average in years, standard deviation | 51.80 ± 13.73 |
Sex | |
Male | 299 (49.7%) |
Female | 302 (50.3%) |
Comorbidities | |
Smoke | 7 (1.16%) |
Arterial Hypertension | 275 (45.9%) |
Diabetes Mellitus | 82 (13.6%) |
Chronic Obstructive Pulmonary Disease | 31 (5.15%) |
Asthma | 54 (9.0%) |
Rhinitis | 60 (10.1%) |
Gastroesophageal Reflux Disease | 57 (9.5%) |
Deviated Septum | 69 (11.5%) |
Heart Failure | 16 (2.6%) |
Coronary Artery Disease | 26 (4.6%) |
Stroke | 13 (2.3%) |
Arrhythmias | 29 (4.9%) |
Epworth Scale | 9.59 ± 4.71 |
Normal < 11 points | 372 (61.8%) |
Very probable Certainty < 16 points | 149 (24.7%) |
Excessive Somnolence ≥ 16 points | 80 (13.5%) |
Body Mass Index—Average in kg/m2 | 30.85 ± 5.78 |
Underweight | 2 (0.3%) |
Normal | 68 (11.3%) |
Overweight | 236 (39.3%) |
Obesity Grade WHO | 295 (49.1%) |
Grade I | 148 (24.6%) |
Grade II | 98 (16.3%) |
Grade III | 49 (8.2%) |
Mallampati | |
Class I | 1 (0.18%) |
Class II | 16 (2.66%) |
Class III | 145 (24.19%) |
Class IV | 439 (73.04%) |
Tonsils Brodsky Classification | |
Grade 0 | 30 (4.99%) |
Grade I | 469 (78.03%) |
Grade II | 71 (11.81%) |
Grade III | 30 (4.99%) |
Grade IV | 1 (0.18%) |
Sleep Latency—min | 24.37 ± 22.01 |
Sleep Effectiveness—average in %, standard deviation | 84.70 ± 38.16 |
Total Sleep Time—average in minutes | 352.39 ± 135.45 |
Microarousal rate—average hourly rate, standard deviation | 27.58 ± 20.35 |
Hypopnea Index—average hourly rate, standard deviation | 22.45 ± 22.00 |
Duration of respiratory events—average in seconds, standard deviation | 22.25 ± 7.42 |
Average oxygen saturation during REM sleep—average in %, standard deviation | 92.71 ± 5.36 |
Average oxygen saturation during NON-REM sleep—average in %, standard deviation | 93.69 ± 2.63 |
Average oxygen saturation during monitoring—average in %, standard deviation | 94.48 ± 2.61 |
Average oxygen saturation during respiratory events—average in %, standard deviation | 89.66 ± 6.23 |
T90—average in %, standard deviation | 7.42 ± 16.71 |
Sleep time with snoring—average in minutes, standard deviation | 11.47 ± 12.53 |
Minimum oxygen saturation during sleep—average in %, standard deviation | 81.11 ± 10.61 |
T90 during REM sleep—average in minutes, standard deviation | 8.73 ± 44.1 |
T90 during Non-REM sleep—average in minutes, standard deviation | 21.51 ± 55.47 |
Hypopnea Index in supine—average in hours, standard deviation | 28.43 ± 26.17 |
IAH in right side—average in hours, standard deviation | 11.93 ± 24.80 |
IAH in left side—average in hours, standard deviation | 1.90 ± 2.47 |
Machine Learning Technique | Accuracy (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | Positive Predictive Value (95% CI) | Negative Predictive Value (95% CI) | Positive Likelihood Ratio (95% CI) | Negative Likelihood Ratio (95% CI) | AUROC |
---|---|---|---|---|---|---|---|---|
Neural Networks | 0.665 (59.7–72.8%) | 0.877 (79.9–93.3%) | 0.453 (35.6–55.2%) | 0.616 (57.1–65.9%) | 0.787 (68.0–86.5%) | 1.60 (1.33–1.93) | 0.27 (0.16–0.47) | 0.665 |
Decision Trees | 0.717 (65.1–77.7%) | 0.906 (83.3–95.4%) | 0.528 (42.9–62.6%) | 0.657 (60.9–70.3%) | 0.848 (75.1–91.2%) | 1.92 (1.56–2.37) | 0.18 (0.10–0.33) | 0.717 |
Random Forests | 0.892 (84.1–93.0%) | 0.943 (88.1–97.9%) | 0.840 (75.6–90.4%) | 0.855 (79.1–90.1%) | 0.937 (87.7–97.0%) | 5.88 (3.80–9.12) | 0.07 (0.03–0.15) | 0.892 |
Extra Trees | 0.896 (84.7–93.4%) | 0.896 (82.2–94.7%) | 0.896 (82.2–94.7%) | 0.896 (83.1–93.8%) | 0.896 (83.1–93.8%) | 8.64 (4.92–15.17) | 0.12 (0.07–0.20) | 0.896 |
ML Technique | Hyperparameters Used during Experimentation | Optimal Hyperparameters Found |
---|---|---|
Neural Networks | Activation functions: identity, logistic, tanh, ReLU | activation: ReLU |
Solvers: Adam, lbfgs, SGD | solver: lbfgs | |
Alpha values: 0 to 1 with increments of 0.1 | alpha: 0.1 | |
Number of hidden layers: 1 to 3 | Number of hidden layers: 2 | |
Number of neurons by layer: 1 to 20 | Hidden_layer_sizes: (6,2) | |
Decision Trees | Class_weight: balanced, None | Class_weight: balanced |
Criterion: entropy, gini | criterion: entropy | |
Max_features: auto, log2, None | splitter: random | |
max_depth: 10 to 200 with increments of 10 | max_depth:120 | |
Random Forests | Criterion: gini, entropy | Criterion: gini, entropy |
n_estimators: 10 to 200 with increments of 10 | n_estimators: 180 | |
min_samples_leaf: 1 to 5 | min_samples_leaf: 1 | |
Extra Trees | Criterion: gini, entropy | criterion: gini |
n_estimators: 10 to 200 with increments of 10 | n_estimators: 20 | |
min_samples_leaf: 1 to 5 | min_samples_leaf: 1 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 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
Bedoya, O.; Rodríguez, S.; Muñoz, J.P.; Agudelo, J. Application of Machine Learning Techniques for the Diagnosis of Obstructive Sleep Apnea/Hypopnea Syndrome. Life 2024, 14, 587. https://doi.org/10.3390/life14050587
Bedoya O, Rodríguez S, Muñoz JP, Agudelo J. Application of Machine Learning Techniques for the Diagnosis of Obstructive Sleep Apnea/Hypopnea Syndrome. Life. 2024; 14(5):587. https://doi.org/10.3390/life14050587
Chicago/Turabian StyleBedoya, Oscar, Santiago Rodríguez, Jenny Patricia Muñoz, and Jared Agudelo. 2024. "Application of Machine Learning Techniques for the Diagnosis of Obstructive Sleep Apnea/Hypopnea Syndrome" Life 14, no. 5: 587. https://doi.org/10.3390/life14050587