Reviewing the connection between speech and obstructive sleep apnea

Abstract

BackgroundSleep apnea (OSA) is a common sleep disorder characterized by recurring breathing pauses during sleep caused by a blockage of the upper airway (UA). The altered UA structure or function in OSA speakers has led to hypothesize the automatic analysis of speech for OSA assessment. In this paper we critically review several approaches using speech analysis and machine learning techniques for OSA detection, and discuss the limitations that can arise when using machine learning techniques for diagnostic applications.MethodsA large speech database including 426 male Spanish speakers suspected to suffer OSA and derived to a sleep disorders unit was used to study the clinical validity of several proposals using machine learning techniques to predict the apnea–hypopnea index (AHI) or classify individuals according to their OSA severity. AHI describes the severity of patients’ condition. We first evaluate AHI prediction using state-of-the-art speaker recognition technologies: speech spectral information is modelled using supervectors or i-vectors techniques, and AHI is predicted through support vector regression (SVR). Using the same database we then critically review several OSA classification approaches previously proposed. The influence and possible interference of other clinical variables or characteristics available for our OSA population: age, height, weight, body mass index, and cervical perimeter, are also studied.ResultsThe poor results obtained when estimating AHI using supervectors or i-vectors followed by SVR contrast with the positive results reported by previous research. This fact prompted us to a careful review of these approaches, also testing some reported results over our database. Several methodological limitations and deficiencies were detected that may have led to overoptimistic results.ConclusionThe methodological deficiencies observed after critically reviewing previous research can be relevant examples of potential pitfalls when using machine learning techniques for diagnostic applications. We have found two common limitations that can explain the likelihood of false discovery in previous research: (1) the use of prediction models derived from sources, such as speech, which are also correlated with other patient characteristics (age, height, sex,…) that act as confounding factors; and (2) overfitting of feature selection and validation methods when working with a high number of variables compared to the number of cases. We hope this study could not only be a useful example of relevant issues when using machine learning for medical diagnosis, but it will also help in guiding further research on the connection between speech and OSA.

Speech Signal and Facial Image Processing for Obstructive Sleep Apnea Assessment

“…Lee et al [ 5 ] do not provide information on the male/female balance in OSA and non-OSA groups, and the significantly lower prevalence of OSA in women compared to men [ 42 ] together with the differences between female and male craniofacial OSA risk factors [ 43 ] may introduce some bias in the performance results. As an illustration of this, in [ 32 ] we have shown how, due to the notable differences between female and male voices [ 44 ], when using speech acoustic features for OSA assessment over OSA and non-OS populations with imbalanced female/male proportions [ 20 ], clearly overoptimistic discrimination results are obtained.…”

“…Furthermore, as the same microphone was used for all recording and all the speakers read the same corpus of four sentences, both channel and phonetic variabilities are minimized so it is reasonable to think that i-vectors will capture characteristics from sounds that can be more affected by OSA. The interested reader may find additional discussion on this topic in [ 32 ] together with a comparison when using supervectors and i-vectors to predict AHI and other clinical variables.…”

Breathing Signature as Vitality Score Index Created by Exercises of Qigong: Implications of Artificial Intelligence Tools Used in Traditional Chinese Medicine

“…Another report came from "using speech analysis and machine learning techniques for obstructive sleep apnea (OSA) detection" for diagnostic applications [50]. "A large speech database including 426 male Spanish speakers suspected to suffer OSA and derived to a sleep disorders unit was used to study the clinical validity of several proposals using machine learning techniques to predict the apnea-hypopnea index (AHI) or classify individuals according to their OSA severity."…”

Obstrüktif Uyku Apnesi Tanıma için Öznitelik Seçimi

“…However, the methods efficiency is bounded by the exploration and exploitation problems of PSO. Espinoza-Cuadros et al explore the relation between the speech signal and the OSA [15]. They use spectral features obtained by i-vectors and train support vector regression (SVR) for the prediction.…”

“…Yilmaz et al, use RR statistical representations of RR intervals as the features and design classifiers using the Support Vector Machines (SVM), Quadratic Discriminant Analysis (QDA), and knearest neighbor (k-NN) [14]. Espinoza-Cuadros et al analysis the machine learning algorithms and features sources for the detection of OSA [15]. Some important conclusions include feature sources (e.g., speech) may be highly correlated with the subject characteristics such as gender and age.…”