Aditif Bayes Ağları Yöntemi ile Obstrüktif Uyku Apnesi Şiddetinin Tahminlenmesi

Öz

Amaç: Günümüzde makine öğrenmesi temelli tahmin modelleri, birçok farklı alanda yaygın bir şekilde kullanılmaktadır. Bu çalışmanın amacı obstrüktif uyku apnesi OUA şiddetinin tahminlenmesinde hastaların klinik özelliklerini içeren bir aditif bayes ağ modeli geliştirilmesidir.Gereç ve Yöntemler: Bu çalışmaya Ocak 2014-Ağustos 2015 tarihleri arasında Akdeniz Üniversitesi Kulak-Burun-Boğaz Hastalıkları Anabilim Dalı’nda OUA ön tanısı ile polisomnografi yapılmış 338 hasta dahil edilmiştir. OUA şiddeti ile ilişkili tüm klinik değişkenler ve bu değişkenlerin birbirleriyle olan ilişkileri aditif bayes ağ modeli ile modellenmiştir.Bulgular: Nihai Bayes modelindeki en önemli öngördürücü değişkenler, yumuşak damak ve uvula boyutu Odds oranı OR :10.94 [1.87 - 64.11] ile mallampati skoru OR: 4.5 [1.46 - 43.22] idi. Nihai modelin 10 kat çapraz geçerlilik test sonucu, OUA şiddetinin 0.632 duyarlılık ve 0.529 seçicilik ile tahminlenebileceğini gösterdi. Ayrıca, geliştirilen modelin ağır OUA hastalarını sınıflandırmada 0.777 duyarlılık ve 0.646 seçiciliğe sahip olduğu saptandı. Sonuç: Makine öğrenmesi temelli tahmin modelleri, klinik değişkenler arasındaki kompleks ilişkileri analiz ederek OUA tanısı ve şiddetinin tahmin edilmesini kolaylaştırabilir

Anahtar Kelimeler

• Aditif Bayes Ağları, Obstruktif uyku apnesi, Klinik karar destek sistemleri

Estimation of Obstructive Sleep Apnea Severity Using Additive Bayesian Networks

Abstract

Objective: Currently, machine learning-based prediction models have been widely used in many different areas. The aim of this study is to develop an additive bayesian network model including characteristics of patients in order to estimate severity of obstructive sleep apnea OSA .Material and Methods: A total of 338 patients who underwent polysomnography due to prediagnosis of OSA at Akdeniz University Department of Otorhinolaryngology between January 2014 and August 2015 were enrolled to this study. All clinical variables related to severity of OSA, and relationships of these variables among each other were modelled by the additive Bayesian network model.Results: The most important predictor variables in the final Bayesian model were the size of soft palate and uvula Odds ratio OR :10.94 [1.87 - 64.11] , and Mallampati score OR: 4.5 [1.46 43.22] . The result of 10-fold cross-validated test for final model indicated that the severity of OSA can be estimated with a sensitivity of 0.632 and specificity of 0.529. In addition, it was determined that the developed model has 0.777 sensitivity and 0.646 specificity to classify severe OSA patients.Conclusion: Machine learning-based prediction models may facilitate the estimation of diagnosis of OSA and its severity by analyzing complex relationships between clinical variables

Keywords

• Additive Bayesian Network, Obstructive Sleep Apnea, Clinical Decision Support Systems

References

1. Jordan AS, McSharry DG, Malhotra A. Adult obstructive sleep apnoea. Lancet 2014; 383:736-47.
2. Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol 2013; 177:1006-14.
3. Young T, Evans L, Finn L, Palta M. Estimation of the clinically diagnosed proportion of sleep apnea syndrome in middle-aged men and women. Sleep 1997; 20:705-6.
4. Bozkurt S, Bostanci A, Turhan M. Can statistical machine learning algorithms help for classification of obstructive sleep apnea severity to optimal utilization of polysomnography resources? Methods Inf Med 2017; 5:308-18.
5. Berry RB, Budhiraja R, Gottlieb DJ, et al. Rules for scoring respiratory events in sleep: Update of the 2007 AASM Manual for the Scoring of Sleep and Associated Events. Deliberations of the Sleep Apnea Definitions Task Force of the American Academy of Sleep Medicine. J Clin Sleep Med 2012; 8: 597-619.
6. Fu LD, Tsamardinos I. A comparison of Bayesian network learning algorithms from continuous data. AMIA Annu Symp Proc 2005: 960.
7. Li J, Zhang C, Wang T, Zhang Y. Generalized Additive Bayesian Network Classifiers. IJCAI 2007; 913-8.
8. World Health Organization. BMI Classification. Global database on body mass index, 2006.

9. Jensen FV. Bayesian Network and Decision Graphs. New York: Springer-Verlag, 2001.
10. Bayat S, Cuggia M, Rossille D, Kessler M, Frimat L. Comparison of Bayesian network and decision tree methods for predicting access to the renal transplant waiting list. Stud Health Technol Inform 2009; 150:600-4.
11. Friedman N, Geiger D, Goldszmidt M. Bayesian Network Classifiers. Machine Learning 1997; 29:131-63.
12. Buntine W. Theory refinement on Bayesian networks. Proceedings of Seventh Conference on Uncertainty in Artificial Intelligence. Los Angeles: Morgan Kaufmann, 1991.
13. Lewis FI, Pittavino M, Furrer R. Data Modelling using Additive Bayesian Networks.
14. Refaeilzadeh P, Tang L, Liu H. Cross-validation. Encyclopedia of database systems: New York: Springer, 2009:532-38.
15. Fushiki T. Estimation of prediction error by using K-fold cross-validation. Statistics and Computing 2009; 21(2):137-46.
16. The R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2006.
17. Leite L, Santos CC, Rodrigues PP. Can we avoid unnecessary polysomnographies in the diagnosis of obstructive sleep apnea? A Bayesian Network Decision support tool. Computer-Based Medical Systems (CBMS), 2014 IEEE 27th International Symposium on; 27-29 May 2014, New York, NY, USA.
18. Musman S, Passos VM, Silva IB, Barreto SM. Evaluation of a prediction model for sleep apnea in patients submitted to polysomnography. J Bras Pneumol 2011; 37:75-84.
19. Rodrigues PP, Santos DF, Leite L. Obstructive sleep apnea diagnosis: The Bayesian Network Model Revisited. Computer-Based Medical Systems (CBMS), 2015 IEEE 28th International Symposium on; 22-25 June 2015, Sao Carlos, Brazil.
20. Sim DYY, Teh CS, Banerjee PK. Prediction model by using Bayesian and Cognition-driven Techniques: A study in the context of obstructive sleep apnea. Procedia - Social and Behavioral Sciences 2013; 97:528-37.