Assessment of oximetry-based statistical classifiers as simplified screening tools in the management of childhood obstructive sleep apnea

Construction and validation of machine learning algorithm for predicting depression among home-quarantined individuals during the large-scale COVID-19 outbreak: based on Adaboost model

OBJECTIVES

COVID-19 epidemics often lead to elevated levels of depression. To accurately identify and predict depression levels in home-quarantined individuals during a COVID-19 epidemic, this study constructed a depression prediction model based on multiple machine learning algorithms and validated its effectiveness.

METHODS

A cross-sectional method was used to examine the depression status of individuals quarantined at home during the epidemic via the network. Characteristics included variables on sociodemographics, COVID-19 and its prevention and control measures, impact on life, work, health and economy after the city was sealed off, and PHQ-9 scale scores. The home-quarantined subjects were randomly divided into training set and validation set according to the ratio of 7:3, and the performance of different machine learning models were compared by 10-fold cross-validation, and the model algorithm with the best performance was selected from 15 models to construct and validate the depression prediction model for home-quarantined subjects. The validity of different models was compared based on accuracy, precision, receiver operating characteristic (ROC) curve, and area under the ROC curve (AUC), and the best model suitable for the data framework of this study was identified.

RESULTS

The prevalence of depression among home-quarantined individuals during the epidemic was 31.66% (202/638), and the constructed Adaboost depression prediction model had an ACC of 0.7917, an accuracy of 0.7180, and an AUC of 0.7803, which was better than the other 15 models on the combination of various performance measures. In the validation sets, the AUC was greater than 0.83.

CONCLUSIONS

The Adaboost machine learning algorithm developed in this study can be used to construct a depression prediction model for home-quarantined individuals that has better machine learning performance, as well as high effectiveness, robustness, and generalizability.

Nocturnal oximetry parameters as predictors of sleep apnea severity in resource-limited settings

Nocturnal oximetry is an alternative modality for evaluating obstructive sleep apnea syndrome (OSAS) severity when polysomnography is not available. The Oxygen Desaturation (≥3%) Index (ODI3) and McGill Oximetry Score (MOS) are used as predictors of moderate-to-severe OSAS (apnea-hypopnea index-AHI >5 episodes/h), an indication for adenotonsillectomy. We hypothesised that ODI3 is a better predictive parameter for AHI >5 episodes/h than the MOS. All polysomnograms performed in otherwise healthy, snoring children with tonsillar hypertrophy in a tertiary hospital (November 2014 to May 2019) were analysed. The ODI3 and MOS were derived from the oximetry channel of each polysomnogram. Logistic regression was applied to assess associations of ODI3 or MOS (predictors) with an AHI >5 episodes/h (primary outcome). Receiver operating characteristic (ROC) curves and areas under ROC curves were used to compare the ODI3 and MOS as predictors of moderate-to-severe OSAS. The optimal cut-off value for each oximetry parameter was determined using Youden's index. Polysomnograms of 112 children (median [interquartile range] age 6.1 [3.9-9.1] years; 35.7% overweight) were analysed. Moderate-to-severe OSAS prevalence was 49.1%. The ODI3 and MOS were significant predictors of moderate-to-severe OSAS after adjustment for overweight, sex, and age (odds ratio [OR] 1.34, 95% confidence interval [CI] 1.19-1.51); and OR 4.10, 95% CI 2.06-8.15, respectively; p < 0.001 for both). Area under the ROC curve was higher for the ODI3 than for MOS (0.903 [95% CI 0.842-0.964] versus 0.745 [95% CI 0.668-0.821]; p < 0.001). Optimal cut-off values for the ODI3 and MOS were ≥4.3 episodes/h and ≥2, respectively. The ODI3 emerges as preferable or at least a complementary oximetry parameter to MOS for detecting moderate-to-severe OSAS in snoring children when polysomnography is not available.

Improving the management of patients with chronic cardiac and respiratory diseases by extending pulse-oximeter uses: the dynamic pulse-oximetry

Respiratory and cardio-vascular chronic diseases are among the most common noncommunicable diseases (NCDs) worldwide, accounting for a large portion of health-care costs in terms of mortality and disability. Their prevalence is expected to rise further in the coming years as the population ages. The current model of care for diagnosing and monitoring NCDs is out of date because it results in late medical interventions and/or an unfavourable cost-effectiveness balance based on reported symptoms and subsequent inpatient tests and treatments. Health projects and programs are being implemented in an attempt to move the time of an NCD's diagnosis, as well as its monitoring and follow up, out of hospital settings and as close to real life as possible, with the goal of benefiting both patients' quality of life and health system budgets. Following the SARS-CoV-2 pandemic, this implementation received additional impetus. Pulseoximeters (POs) are currently used in a variety of clinical settings, but they can also aid in the telemonitoring of certain patients. POs that can measure activities as well as pulse rate and oxygen saturation as proxies of cardio-vascular and respiratory function are now being introduced to the market. To obtain these data, the devices must be absolutely reliable, that is, accurate and precise, and capable of recording for a long enough period of time to allow for diagnosis. This paper is a review of current pulse-oximetry (POy) use, with the goal of investigating how its current use can be expanded to manage not only cardio-respiratory NCDs, but also acute emergencies with telemonitoring when hospitalization is not required but the patients' situation is debatable. Newly designed devices, both "consumer" and "professional," will be scrutinized, particularly those capable of continuously recording vital parameters on a 24-hour basis and coupling them with daily activities, a practice known as dynamic pulse-oximetry.

Reliability of machine learning to diagnose pediatric obstructive sleep apnea: Systematic review and meta-analysis

BACKGROUND

Machine-learning approaches have enabled promising results in efforts to simplify the diagnosis of pediatric obstructive sleep apnea (OSA). A comprehensive review and analysis of such studies increase the confidence level of practitioners and healthcare providers in the implementation of these methodologies in clinical practice.

OBJECTIVE

To assess the reliability of machine-learning-based methods to detect pediatric OSA.

DATA SOURCES

Two researchers conducted an electronic search on the Web of Science and Scopus using term, and studies were reviewed along with their bibliographic references.

ELIGIBILITY CRITERIA

Articles or reviews (Year 2000 onwards) that applied machine learning to detect pediatric OSA; reported data included information enabling derivation of true positive, false negative, true negative, and false positive cases; polysomnography served as diagnostic standard.

APPRAISAL AND SYNTHESIS METHODS

Pooled sensitivities and specificities were computed for three apnea-hypopnea index (AHI) thresholds: 1 event/hour (e/h), 5 e/h, and 10 e/h. Random-effect models were assumed. Summary receiver-operating characteristics (SROC) analyses were also conducted. Heterogeneity (I ² ) was evaluated, and publication bias was corrected (trim and fill).

RESULTS

Nineteen studies were finally retained, involving 4767 different pediatric sleep studies. Machine learning improved diagnostic performance as OSA severity criteria increased reaching optimal values for AHI = 10 e/h (0.652 sensitivity; 0.931 specificity; and 0.940 area under the SROC curve). Publication bias correction had minor effect on summary statistics, but high heterogeneity was observed among the studies.

Oximetry Indices in the Management of Sleep Apnea: From Overnight Minimum Saturation to the Novel Hypoxemia Measures

Obstructive sleep apnea (OSA) is a multidimensional disease often underdiagnosed due to the complexity and unavailability of its standard diagnostic method: the polysomnography. Among the alternative abbreviated tests searching for a compromise between simplicity and accurateness, oximetry is probably the most popular. The blood oxygen saturation (SpO₂) signal is characterized by a near-constant profile in healthy subjects breathing normally, while marked drops (desaturations) are linked to respiratory events. Parameterization of the desaturations has led to a great number of indices of severity assessment commonly used to assist in OSA diagnosis. In this chapter, the main methodologies used to characterize the overnight oximetry profile are reviewed, from visual inspection and simple statistics to complex measures involving signal processing and pattern recognition techniques. We focus on the individual performance of each approach, but also on the complementarity among the great amount of indices existing in the state of the art, looking for the most relevant oximetric feature subset. Finally, a quick overview of SpO₂-based deep learning applications for OSA management is carried out, where the raw oximetry signal is analyzed without previous parameterization. Our research allows us to conclude that all the methodologies (conventional, time, frequency, nonlinear, and hypoxemia-based) demonstrate high ability to provide relevant oximetric indices, but only a reduced set provide non-redundant complementary information leading to a significant performance increase. Finally, although oximetry is a robust tool, greater standardization and prospective validation of the measures derived from complex signal processing techniques are still needed to homogenize interpretation and increase generalizability.

Diagnostic approaches to respiratory abnormalities in craniofacial syndromes

Craniofacial syndromes are a complex cluster of genetic conditions characterized by embryonic perturbations in the developmental trajectory of the upper airway and related structures. The presence of reduced airway size and maladaptive neuromuscular responses, particularly during sleep, leads to significant alterations in sleep architecture and overall detrimental gas exchange abnormalities that can be life-threatening. The common need for multi-stage therapeutic interventions for these craniofacial problems requires careful titration of anatomy and function, and the latter is currently evaluated by overnight polysomnography in sleep laboratories. The cost, inconvenience, and scarcity of pediatric sleep laboratories preclude the frequent evaluations that could optimize the overall process of treatment and corresponding outcomes. Here, we critically examine reductionist approaches to polysomnography in children to establish the parallel approximation of such techniques to infant with craniofacial disorders. The need for prospective longitudinal multicenter studies with side-by-side comparisons aimed at identifying an optimal diagnostic and long-term monitoring paradigm for these potentially life-threatening conditions is emphasized.

Diagnostic accuracy of level IV portable sleep monitors versus polysomnography for pediatric obstructive sleep apnea: a systematic review and meta-analysis

BACKGROUND

Obstructive sleep apnea (OSA) is one of the common sleep-related breathing disorders in children. However, polysomnography (PSG) is an expensive and labor-intensive diagnostic modality that may not always be feasible, especially in low-income countries or in non-tertiary hospitals. Portable monitors (PMs), a new approach for OSA diagnosis, have become more widely used with lower intolerance and cost in recent years. We aimed to analyze the diagnostic performance of Level IV PMs compared with PSG for the diagnosis of pediatric OSA.

METHODS

PubMed and Embase databases were searched for studies published in English up to December 31, 2020 evaluating the diagnostic accuracy of Level IV PMs against the apnea-hypopnea index (AHI) measured using overnight in-laboratory polysomnography (PSG) in children and adolescents. A random-effects bivariate model was used to estimate the summary sensitivity and specificity of oximetry-based statistical classifiers. A qualitative evaluation of studies was performed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) rating.

RESULTS

In total, 20 studies involving 7062 participants were included in this systematic review. Among these articles, seven studies (oximetry based on new mathematical classifiers) involving 5098 individuals satisfied the criteria for quantitative synthesis. Compared with AHI evaluation measured by PSG, different PM systems achieved diagnostic accuracy with variable degrees of success. A meta-analysis showed a pooled sensitivity of 74% (95% confidence interval [CI]: 66-80%) and pooled specificity of 90% (95% CI: 85-94%). The area under the summary receiver operating characteristic (SROC) curve was 0.89 (95% CI: 0.86-0.92).

CONCLUSION

This study showed the potential of Level IV PMs for screening pediatric OSA patients. Oximetry based on new mathematical classifiers may provide a simple and effective alternative to PSG in the diagnosis of pediatric OSA especially in the context of appropriate clinical evaluation.

Alternatives to surgery in children with mild OSA

Precision medicine requires coordinated and integrated evidence-based combinatorial approaches so that diagnosis and treatment can be tailored to the individual patient. In this context, the treatment approach to mild obstructive sleep apnea (OSA) is fraught with substantial debate as to what is mild OSA, and as to what constitutes appropriate treatment. As such, it is necessary to first establish a proposed consensus of what criteria need to be employed to reach the diagnosis of mild OSA, and then examine the circumstances under which treatment is indicated, and if so, whether and when anti-inflammatory therapy (AIT), rapid maxillary expansion (RME), and/or myofunctional therapy (MFT) may be indicated.

A Convolutional Neural Network Architecture to Enhance Oximetry Ability to Diagnose Pediatric Obstructive Sleep Apnea

This study aims at assessing the usefulness of deep learning to enhance the diagnostic ability of oximetry in the context of automated detection of pediatric obstructive sleep apnea (OSA). A total of 3196 blood oxygen saturation (SpO2) signals from children were used for this purpose. A convolutional neural network (CNN) architecture was trained using 20-min SpO2 segments from the training set (859 subjects) to estimate the number of apneic events. CNN hyperparameters were tuned using Bayesian optimization in the validation set (1402 subjects). This model was applied to three test sets composed of 312, 392, and 231 subjects from three independent databases, in which the apnea-hypopnea index (AHI) estimated for each subject (AHICNN) was obtained by aggregating the output of the CNN for each 20-min SpO2 segment. AHICNN outperformed the 3% oxygen desaturation index (ODI3), a clinical approach, as well as the AHI estimated by a conventional feature-engineering approach based on multi-layer perceptron (AHIMLP). Specifically, AHICNN reached higher four-class Cohen's kappa in the three test databases than ODI3 (0.515 vs 0.417, 0.422 vs 0.372, and 0.423 vs 0.369) and AHIMLP (0.515 vs 0.377, 0.422 vs 0.381, and 0.423 vs 0.306). In addition, our proposal outperformed state-of-the-art studies, particularly for the AHI severity cutoffs of 5 e/h and 10 e/h. This suggests that the information automatically learned from the SpO2 signal by deep-learning techniques helps to enhance the diagnostic ability of oximetry in the context of pediatric OSA.

Development of a Minimally Invasive Screening Tool to Identify Obese Pediatric Population at Risk of Obstructive Sleep Apnea/Hypopnea Syndrome

Obstructive sleep apnea syndrome is a reduction of the airflow during sleep which not only produces a reduction in sleep quality but also has major health consequences. The prevalence in the obese pediatric population can surpass 50%, and polysomnography is the current gold standard method for its diagnosis. Unfortunately, it is expensive, disturbing and time-consuming for experienced professionals. The objective is to develop a patient-friendly screening tool for the obese pediatric population to identify those children at higher risk of suffering from this syndrome. Three supervised learning classifier algorithms (i.e., logistic regression, support vector machine and AdaBoost) common in the field of machine learning were trained and tested on two very different datasets where oxygen saturation raw signal was recorded. The first dataset was the Childhood Adenotonsillectomy Trial (CHAT) consisting of 453 individuals, with ages between 5 and 9 years old and one-third of the patients being obese. Cross-validation was performed on the second dataset from an obesity assessment consult at the Pediatric Department of the Hospital General Universitario of Valencia. A total of 27 patients were recruited between 5 and 17 years old; 42% were girls and 63% were obese. The performance of each algorithm was evaluated based on key performance indicators (e.g., area under the curve, accuracy, recall, specificity and positive predicted value). The logistic regression algorithm outperformed (accuracy = 0.79, specificity = 0.96, area under the curve = 0.9, recall = 0.62 and positive predictive value = 0.94) the support vector machine and the AdaBoost algorithm when trained with the CHAT datasets. Cross-validation tests, using the Hospital General de Valencia (HG) dataset, confirmed the higher performance of the logistic regression algorithm in comparison with the others. In addition, only a minor loss of performance (accuracy = 0.75, specificity = 0.88, area under the curve = 0.85, recall = 0.62 and positive predictive value = 0.83) was observed despite the differences between the datasets. The proposed minimally invasive screening tool has shown promising performance when it comes to identifying children at risk of suffering obstructive sleep apnea syndrome. Moreover, it is ideal to be implemented in an outpatient consult in primary and secondary care.

Predicting polysomnographic severity thresholds in children using machine learning

BACKGROUND

Approximately 500,000 children undergo tonsillectomy and adenoidectomy (T&A) annually for treatment of obstructive sleep disordered breathing (oSDB). Although polysomnography is beneficial for preoperative risk stratification in these children, its expanded use is limited by the associated costs and resources needed. Therefore, we used machine learning and data from potentially wearable sensors to identify children needing postoperative overnight monitoring based on the polysomnographic severity of oSDB.

METHODS

Children aged 2-17 years undergoing polysomnography were included. Six machine learning models were created using (i) clinical parameters and (ii) nocturnal actigraphy and oxygen desaturation index. The prediction performance for polysomnography-derived severity of oSDB measured by apnea hypopnea index (AHI) >2 and >10 were evaluated.

RESULTS

One hundred and ninety children were included. One hundred and eight were male (57%), mean age was 6.7 years [95% confidence interval; 6.1, 7.2], and mean AHI was 10.6 [7.8, 13.4]. Predictive performance utilizing clinical parameters was poor for both AHI > 2 (accuracy range: 48-56% for all models) and AHI > 10 (50-61%). Combining oximetry and actigraphy improved the accuracy to 87-89% for AHI > 2 and 95-96% for AHI > 10.

CONCLUSIONS

Machine learning with oximetry and actigraphy identifies most children needing overnight monitoring as determined by polysomnographic severity of oSDB, supporting a potential resource-conscious screening pathway for children undergoing T&A.

IMPACT

We provide proof of principle for the utility of machine learning, oximetry, and actigraphy to screen for severe obstructive sleep apnea syndrome (OSAS) in children. Clinical parameters perform poorly in predicting the severity of OSAS, which is confirmed in the current study. The predictive accuracy for severe OSAS was improved by a smaller subset of quantifiable physiologic parameters, such as oximetry. The results of this study support a lower cost, patient-friendly screening pathway to identify children in need of in-hospital observation after surgery.

Assessment of Airflow and Oximetry Signals to Detect Pediatric Sleep Apnea-Hypopnea Syndrome Using AdaBoost

The reference standard to diagnose pediatric Obstructive Sleep Apnea (OSA) syndrome is an overnight polysomnographic evaluation. When polysomnography is either unavailable or has limited availability, OSA screening may comprise the automatic analysis of a minimum number of signals. The primary objective of this study was to evaluate the complementarity of airflow (AF) and oximetry (SpO₂) signals to automatically detect pediatric OSA. Additionally, a secondary goal was to assess the utility of a multiclass AdaBoost classifier to predict OSA severity in children. We extracted the same features from AF and SpO₂ signals from 974 pediatric subjects. We also obtained the 3% Oxygen Desaturation Index (ODI) as a common clinically used variable. Then, feature selection was conducted using the Fast Correlation-Based Filter method and AdaBoost classifiers were evaluated. Models combining ODI 3% and AF features outperformed the diagnostic performance of each signal alone, reaching 0.39 Cohens's kappa in the four-class classification task. OSA vs. No OSA accuracies reached 81.28%, 82.05% and 90.26% in the apnea-hypopnea index cutoffs 1, 5 and 10 events/h, respectively. The most relevant information from SpO₂ was redundant with ODI 3%, and AF was complementary to them. Thus, the joint analysis of AF and SpO₂ enhanced the diagnostic performance of each signal alone using AdaBoost, thereby enabling a potential screening alternative for OSA in children.

Artificial intelligence in sleep medicine: background and implications for clinicians

None

Polysomnography remains the cornerstone of objective testing in sleep medicine and results in massive amounts of electrophysiological data, which is well-suited for analysis with artificial intelligence (AI)-based tools. Combined with other sources of health data, AI is expected to provide new insights to inform the clinical care of sleep disorders and advance our understanding of the integral role sleep plays in human health. Additionally, AI has the potential to streamline day-to-day operations and therefore optimize direct patient care by the sleep disorders team. However, clinicians, scientists, and other stakeholders must develop best practices to integrate this rapidly evolving technology into our daily work while maintaining the highest degree of quality and transparency in health care and research. Ultimately, when harnessed appropriately in conjunction with human expertise, AI will improve the practice of sleep medicine and further sleep science for the health and well-being of our patients.

Cloud algorithm-driven oximetry-based diagnosis of obstructive sleep apnoea in symptomatic habitually snoring children

The ability of a cloud-driven Bluetooth oximetry-based algorithm to diagnose obstructive sleep apnoea syndrome (OSAS) was examined in habitually snoring children concurrently undergoing overnight polysomnography.

Children clinically referred for overnight in-laboratory polysomnographic evaluation for suspected OSAS were simultaneously hooked to a Bluetooth oximeter linked to a smartphone. Polysomnography findings were scored and the apnoea/hypopnoea index (AHIPSG) was tabulated, while oximetry data yielded an estimated AHIOXI using a validated algorithm.

The accuracy of the oximeter in identifying correctly patients with OSAS in general, or with mild (AHI 1–5 events·h−1), moderate (5–10 events·h−1) or severe (>10 events·h−1) OSAS was examined in 432 subjects (6.5±3.2 years), with 343 having AHIPSG >1 event·h−1. The accuracies of AHIOXI were consistently >79% for all levels of OSAS severity, and specificity was particularly favourable for AHI >10 events·h−1 (92.7%). Using the criterion of AHIPSG >1 event·h−1, only 4.7% of false-negative cases emerged, from which only 0.6% of cases showed moderate or severe OSAS.

Overnight oximetry processed via Bluetooth technology by a cloud-based machine learning-derived algorithm can reliably diagnose OSAS in children with clinical symptoms suggestive of the disease. This approach provides virtually limitless scalability and should alleviate the substantial difficulties in accessing paediatric sleep laboratories while markedly reducing the costs of OSAS diagnosis.

Pediatric pulse oximetry-based OSA screening at different thresholds of the apnea-hypopnea index with an expression of uncertainty for inconclusive classifications

BACKGROUND

Assessments of pediatric obstructive sleep apnea (OSA) are underutilized across Canada due to a lack of resources. Polysomnography (PSG) measures OSA severity through the average number of apnea/hypopnea events per hour (AHI), but is resource intensive and requires a specialized sleep laboratory, which results in long waitlists and delays in OSA detection. Prompt diagnosis and treatment of OSA are crucial for children, as untreated OSA is linked to behavioral deficits, growth failure, and negative cardiovascular consequences. We aim to assess the performance of a portable pediatric OSA screening tool at different AHI cut-offs using overnight smartphone-based pulse oximetry.

MATERIAL AND METHODS

Following ethics approval and informed consent, children referred to British Columbia Children's Hospital for overnight PSG were recruited for two studies including 160 and 75 children, respectively. An additional smartphone-based pulse oximeter sensor was used in both studies to record overnight pulse oximetry [SpO₂ and photoplethysmogram (PPG)] alongside the PSG. Features characterizing SpO₂ dynamics and heart rate variability from pulse peak intervals of the PPG signal were derived from pulse oximetry recordings. Three multivariate logistic regression screening models, targeted at three different levels of OSA severity (AHI ≥ 1, 5, and 10), were developed using stepwise-selection of features using the Bayesian information criterion (BIC). The "Gray Zone" approach was also implemented for different tolerance values to allow for more precise detection of children with inconclusive classification results.

RESULTS

The optimal diagnostic tolerance values defining the "Gray Zone" borders (15, 10, and 5, respectively) were selected to develop the final models to screen for children at AHI cut-offs of 1, 5, and 10. The final models evaluated through cross-validation showed good accuracy (75%, 82% and 89%), sensitivity (80%, 85% and 82%) and specificity (65%, 79% and 91%) values for detecting children with AHI ≥ 1, AHI ≥ 5 and AHI ≥ 10. The percentage of children classified as inconclusive was 28%, 38% and 16% for models detecting AHI ≥ 1, AHI ≥ 5, and AHI ≥ 10, respectively.

CONCLUSIONS

The proposed pulse oximetry-based OSA screening tool at different AHI cut-offs may assist clinicians in identifying children at different OSA severity levels. Using this tool at home prior to PSG can help with optimizing the limited resources for PSG screening. Further validation with larger and more heterogeneous datasets is required before introducing in clinical practice.