Sleep apnea detection from single-lead electrocardiogram signals using effective deep-shallow fusion network

Objective.Explore a network architecture that can efficiently perform single-lead electrocardiogram (ECG) sleep apnea (SA) detection by utilizing the beneficial information of extended ECG segments and reducing the impact of their noisy information.Approach.We propose an effective deep-shallow fusion network (EDSFnet). The deeper residual network is used to extract high-level features with stronger semantics and less noise from the original ECG segments. The shallower convolutional neural network is used to extract lower-level features with higher resolution containing more detailed neighborhood information from the extended ECG segments. These two types of features are then fused using Effective Channel Attention, implementing automatic weight assignment to take advantage of their complementary nature.Main results.The performance of EDSFnet is evaluated on the Apnea-ECG dataset and the FAH-ECG dataset. In the Apnea-ECG dataset with 35 subjects as the training set and 35 subjects as the test set, the accuracy of EDSFnet was 92.6% and 100% for per-segment and per-recording test, respectively. In the FAH-ECG dataset with 348 subjects as the training set and 88 subjects as the test set, the accuracy of EDSFnet was 89.0% and 93.2% for per-segment and per-recording test, respectively. EDSFnet has achieved state-of-the-art results in both experiments using the publicly available Apnea-ECG dataset and subject-independent experiments using the FAH-ECG clinical dataset.Significance.The success of EDSFnet in handling SA detection underlines its robustness and adaptability. By achieving superior results across different datasets, EDSFnet offers promise in advancing the cost-effective and efficient detection of SA through single-lead ECG, reducing the burden on patients and healthcare systems alike.

Prevalence of Obstructive Sleep Apnea Using Home Sleep Test in Taiwan During the Coronavirus Disease Pandemic

Background

Obstructive sleep apnea syndrome (OSAS) is a common disorder associated with serious sequelae. The current gold standard diagnostic method, polysomnography, is costly and time consuming and requires patients to stay overnight at a facility.

Aim

This study aimed to reveal the prevalence of OSAS in general adult population using a home sleep test (HST) during the coronavirus disease 2019 (COVID-19) pandemic.

Methods

This prospective cohort study was conducted by the Department of Otolaryngology, Taipei City Hospital, Taipei, Taiwan, between January 2020 and December 2021. A total of 1372 patients aged 30-70 years completed an HST using a Type 3 portable sleep monitor (PM). The apnea-hypopnea index (AHI) was analyzed to assess the association of OSAS with age, body mass index (BMI), sex, Epworth Sleepiness Scale (ESS) and the Sleep Apnea Risk Assessment questionnaire (STOP-Bang questionnaire) rating.

Results

The mean age of the patients (782 men, 57%; 590 women, 43%) was 49.24 ± 11.04 years. OSAS was detected in 954 (69.5%) patients with 399 (29.1%) mild OSAS; 246 (17.9%) moderate OSAS; and 309 (22.5%) severe OSAS. Among these, the prevalence of moderate-to-severe OSAS was 143 (10.4%) in women and 412 (30.0%) in men. The mean age was the highest (51.29 ± 11.29) in the mild OSAS group and lowest (47.08 ± 10.87) in the healthy group. OSAS severity was greater with increasing BMI, 23.39 ± 3.44 in the healthy group and 29.29 ± 5.01 in the severe OSAS group. A positive correlation was also noted between the ESS/STOP-Bang questionnaire rating and OSAS severity.

Conclusion

The prevalence of OSAS in Taiwan was 69.5% in our study. It showed strong evidence that OSAS has important public health consequences and PMs are simple, fast, feasible, and cost-effective tools for OSAS screening in the home environment, especially during the COVID-19 pandemic.

A deep learning-based algorithm for detection of cortical arousal during sleep

STUDY OBJECTIVES

The frequency of cortical arousals is an indicator of sleep quality. Additionally, cortical arousals are used to identify hypopneic events. However, it is inconvenient to record electroencephalogram (EEG) data during home sleep testing. Fortunately, most cortical arousal events are associated with autonomic nervous system activity that could be observed on an electrocardiography (ECG) signal. ECG data have lower noise and are easier to record at home than EEG. In this study, we developed a deep learning-based cortical arousal detection algorithm that uses a single-lead ECG to detect arousal during sleep.

METHODS

This study included 1,547 polysomnography records that met study inclusion criteria and were selected from the Multi-Ethnic Study of Atherosclerosis database. We developed an end-to-end deep learning model consisting of convolutional neural networks and recurrent neural networks which: (1) accepted varying length physiological data; (2) directly extracted features from the raw ECG signal; (3) captured long-range dependencies in the physiological data; and (4) produced arousal probability in 1-s resolution.

RESULTS

We evaluated the model on a test set (n = 311). The model achieved a gross area under precision-recall curve score of 0.62 and a gross area under receiver operating characteristic curve score of 0.93.

CONCLUSION

This study demonstrated the end-to-end deep learning approach with a single-lead ECG has the potential to be used to accurately detect arousals in home sleep tests.

Continuous Vital Monitoring During Sleep and Light Activity Using Carbon-Black Elastomer Sensors

The comfortable, continuous monitoring of vital parameters is still a challenge. The long-term measurement of respiration and cardiovascular signals is required to diagnose cardiovascular and respiratory diseases. Similarly, sleep quality assessment and the recovery period following acute treatments require long-term vital parameter datalogging. To address these requirements, we have developed “VitalCore”, a wearable continuous vital parameter monitoring device in the form of a T-shirt targeting the uninterrupted monitoring of respiration, pulse, and actigraphy. VitalCore uses polymer-based stretchable resistive bands as the primary sensor to capture breathing and pulse patterns from chest expansion. The carbon black-impregnated polymer is implemented in a U-shaped configuration and attached to the T-shirt with “interfacing” material along with the accompanying electronics. In this paper, VitalCore is bench tested and compared to gold standard respiration and pulse measurements to verify its functionality and further to assess the quality of data captured during sleep and during light exercise (walking). We show that these polymer-based sensors could identify respiratory peaks with a sensitivity of 99.44%, precision of 96.23%, and false-negative rate of 0.557% during sleep. We also show that this T-shirt configuration allows the wearer to sleep in all sleeping positions with a negligible difference of data quality. The device was also able to capture breathing during gait with 88.9–100% accuracy in respiratory peak detection.

Predictors of sleep-disordered breathing in pregnancy

BACKGROUND

Sleep-disordered breathing (SDB) is common in pregnancy, but there are limited data on predictors.

OBJECTIVES

The objective of this study was to develop predictive models of sleep-disordered breathing during pregnancy.

STUDY DESIGN

Nulliparous women completed validated questionnaires to assess for symptoms related to snoring, fatigue, excessive daytime sleepiness, insomnia, and restless leg syndrome. The questionnaires included questions regarding the timing of sleep and sleep duration, work schedules (eg, shift work, night work), sleep positions, and previously diagnosed sleep disorders. Frequent snoring was defined as self-reported snoring ≥3 days per week. Participants underwent in-home portable sleep studies for sleep-disordered breathing assessment in early (6-15 weeks gestation) and mid pregnancy (22-31 weeks gestation). Sleep-disordered breathing was characterized by an apnea hypopnea index that included all apneas, plus hypopneas with ≥3% oxygen desaturation. For primary analyses, an apnea hypopnea index ≥5 events per hour was used to define sleep-disordered breathing. Odds ratios and 95% confidence intervals were calculated for predictor variables. Predictive ability of the logistic models was estimated with area under the receiver-operating-characteristic curves, along with sensitivities, specificities, and positive and negative predictive values and likelihood ratios.

RESULTS

Among 3705 women who were enrolled, data were available for 3264 and 2512 women in early and mid pregnancy, respectively. The corresponding prevalence of sleep-disordered breathing was 3.6% and 8.3%, respectively. At each time point in gestation, frequent snoring, chronic hypertension, greater maternal age, body mass index, neck circumference, and systolic blood pressure were associated most strongly with an increased risk of sleep-disordered breathing. Logistic regression models that included current age, body mass index, and frequent snoring predicted sleep-disordered breathing in early pregnancy, sleep-disordered breathing in mid pregnancy, and new onset sleep-disordered breathing in mid pregnancy with 10-fold cross-validated area under the receiver-operating-characteristic curves of 0.870, 0.838, and 0.809. We provide a supplement with expanded tables, integrated predictiveness, classification curves, and an predicted probability calculator.

CONCLUSION

Among nulliparous pregnant women, logistic regression models with just 3 variables (ie, age, body mass index, and frequent snoring) achieved good prediction of prevalent and incident sleep-disordered breathing. These results can help with screening for sleep-disordered breathing in the clinical setting and for future clinical treatment trials.

Watch Peripheral Arterial Tonometry in the Diagnosis of Pediatric Obstructive Sleep Apnea

Objective To assess the accuracy and clinical reliability of watch peripheral arterial tonometry (PAT) compared with polysomnography (PSG) for the diagnosis of pediatric obstructive sleep apnea (OSA). Study Design Prospective, diagnostic test study. Setting National tertiary referral hospital. Subjects and Methods Patients aged 8 to 15 years with clinically suspected OSA were recruited. All participants underwent PSG and PAT simultaneously in the sleep laboratory. Results Thirty-six patients were included, with a mean age of 10.2 ± 1.8 years. Median (interquartile range) of the apnea hypopnea index (AHI) was 8.0 (5.5-12) and 2.9 (0.5-7.5) events/h, median oxygen desaturation index (ODI) was 2.5 (1.4-8.3) and 1.3 (0.2-3.8) events/h, mean ± standard deviation total sleep time was 398.4 ± 38.3 and 401.9 ± 36.1 minutes, and mean minimum oxygen saturation was 87.1% ± 8.1% and 89.4% ± 7.1% for PSG and PAT sleep parameter results, respectively. Agreement between methods was excellent for the AHI (intraclass correlation coefficient [ICC]: 0.89; 95% CI, 0.40-0.96; P < .001) and ODI (ICC: 0.87; 95% CI, 0.69-0.94; P < .001). Correlation between methods was very good for the ODI ( r = 0.83; 95% CI, 0.67-0.90; P < .001) and moderate for the AHI ( r = 0.64; 95% CI, 0.30-0.85; P < .001). From the receiver operating characteristic curve constructed to assess PAT diagnostic capability, AHI of PAT (W-AHI) at a cutoff of 3.5 events/h provided the highest accuracy (76.9% sensitivity, 78.3% specificity), while W-AHI at 10 events/h yielded 91.3% specificity for diagnosing severe OSA. Conclusion PAT correlated well and had good agreement with PSG. Children with W-AHI ≥10 had high specificity for the diagnosis of severe OSA. Larger studies with PAT designed for children across all age ranges and with a normal control group are still needed.

A Feedback-Controlled Mandibular Positioner Identifies Individuals With Sleep Apnea Who Will Respond to Oral Appliance Therapy

STUDY OBJECTIVES

Mandibular protruding oral appliances represent a potentially important therapy for obstructive sleep apnea (OSA). However, their clinical utility is limited by a less-than-ideal efficacy rate and uncertainty regarding an efficacious mandibular position, pointing to the need for a tool to assist in delivery of the therapy. The current study assesses the ability to prospectively identify therapeutic responders and determine an efficacious mandibular position.

METHODS

Individuals (n = 202) with OSA participated in a blinded, 2-part investigation. A system for identifying therapeutic responders was developed in part 1 (n = 149); the predictive accuracy of this system was prospectively evaluated on a new population in part 2 (n = 53). Each participant underwent a 2-night, in-home feedback-controlled mandibular positioner (FCMP) test, followed by treatment with a custom oral appliance and an outcome study with the oral appliance in place. A machine learning classification system was trained to predict therapeutic outcome on data obtained from FCMP studies on part 1 participants. The accuracy of this trained system was then evaluated on part 2 participants by examining the agreement between prospectively predicted outcome and observed outcome. A predicted efficacious mandibular position was derived from each FCMP study.

RESULTS

Predictive accuracy was as follows: sensitivity 85%; specificity 93%; positive predictive value 97%; and negative predictive value 72%. Of participants correctly predicted to respond to therapy, the predicted mandibular protrusive position proved efficacious in 86% of cases.

CONCLUSIONS

An unattended, in-home FCMP test prospectively identifies individuals with OSA who will respond to oral appliance therapy and provides an efficacious mandibular position.

CLINICAL TRIAL REGISTRATION

The trial that this study reports on is registered on www.clinicaltrials.gov, ID NCT03011762, study name: Feasibility and Predictive Accuracy of an In-Home Computer Controlled Mandibular Positioner in Identifying Favourable Candidates for Oral Appliance Therapy.

Decision Modeling in Sleep Apnea: The Critical Roles of Pretest Probability, Cost of Untreated Obstructive Sleep Apnea, and Time Horizon

STUDY OBJECTIVES

Obstructive sleep apnea (OSA) is associated with increased morbidity and mortality, and treatment with positive airway pressure (PAP) is cost-effective. However, the optimal diagnostic strategy remains a subject of debate. Prior modeling studies have not consistently supported the widely held assumption that home sleep testing (HST) is cost-effective.

METHODS

We modeled four strategies: (1) treat no one; (2) treat everyone empirically; (3) treat those testing positive during in-laboratory polysomnography (PSG) via in-laboratory titration; and (4) treat those testing positive during HST with auto-PAP. The population was assumed to lack independent reasons for in-laboratory PSG (such as insomnia, periodic limb movements in sleep, complex apnea). We considered the third-party payer perspective, via both standard (quality-adjusted) and pure cost methods.

RESULTS

The preferred strategy depended on three key factors: pretest probability of OSA, cost of untreated OSA, and time horizon. At low prevalence and low cost of untreated OSA, the treat no one strategy was favored, whereas empiric treatment was favored for high prevalence and high cost of untreated OSA. In-laboratory backup for failures in the at-home strategy increased the preference for the at-home strategy. Without laboratory backup in the at-home arm, the in-laboratory strategy was increasingly preferred at longer time horizons.

CONCLUSION

Using a model framework that captures a broad range of clinical possibilities, the optimal diagnostic approach to uncomplicated OSA depends on pretest probability, cost of untreated OSA, and time horizon. Estimating each of these critical factors remains a challenge warranting further investigation.

Sex differences in sleep apnea predictors and outcomes from home sleep apnea testing

STUDY OBJECTIVES

To evaluate sex differences in predictors of obstructive sleep apnea (OSA) as per outcomes from home sleep apnea testing.

DESIGN

This was a retrospective analysis of a large repository of anonymous test results and pretest risk factors for OSA.

SETTING AND PATIENTS

A total of 272,705 patients were referred for home sleep apnea testing from a variety of clinical practices for suspected sleep disordered breathing across North America from 2009 to 2013.

INTERVENTIONS

Not applicable.

MEASUREMENTS AND RESULTS

Predictors of OSA (apnea hypopnea index4%≥5) were evaluated by multiple logistic regression; sex differences were evaluated by interaction effects. Middle age was the single most robust predictor of OSA for both sexes and was particularly foretelling for females (P<0.001) even after controlling for measures of adiposity and medical conditions. Females over the age of 45 years were much more likely to have OSA compared to their younger counterparts (78.7% vs 42.5%, respectively; odds ratio: 5.0) versus males (88.1% vs 68.8%, respectively; odds ratio: 3.4). Snoring, although more frequently reported by males, was similarly predictive of OSA for both sexes. Witnessed apneas and measures of adiposity were better predictors of OSA for males than females. Insomnia, depression, and use of sleep medication, although more commonly reported in females, did not predict OSA. Hypertension, although equally reported by both sexes, performed better as a predictor in females (P<0.001), even after controlling for age, measures of adiposity, and other medical conditions. Diabetes, heart disease, stroke, and sleepiness did not contribute unique variance in OSA in adjusted models.

CONCLUSION

This study found that males and females report different symptoms upon clinical evaluation for suspected sleep apnea, with some of the "classic" OSA features to be more common in and robustly predictive for males. The finding that advancing age uniquely and robustly predicted OSA in females reinforces our understanding that age-related changes in sex hormones play a role in the development and/or manifestation of sleep disordered breathing. Need exists for sex-specific prediction models and quantification of menopausal status in OSA screening tools.

Predictors of Obstructive Sleep Apnea on Polysomnography after a Technically Inadequate or Normal Home Sleep Test

STUDY OBJECTIVES

Home sleep testing (HST) is an accepted alternative to polysomnography (PSG) for diagnosing obstructive sleep apnea (OSA) in high-risk populations. Clinical guidelines recommend PSG in cases where the HST is technically inadequate (TI) or fails to establish the diagnosis of OSA in patients with high pretest probability. This retrospective study evaluated predictors of OSA on PSG within patients who had a TI or normal HST.

METHODS

Electronic medical records were reviewed on 1,157 patients referred for HST at our sleep center. Two hundred thirty-eight patients had a TI or normal HST with subsequent PSG. Age, BMI, Epworth score, HST result, and PSG-based apnea-hypopnea index (AHI) were abstracted.

RESULTS

Two hundred thirty-eight consecutive patients with either a normal HST (n = 127) or TI HST (n = 111) underwent subsequent PSG. Of 127 who had a normal HST, 76% had a normal PSG and 24% had OSA (23 mild, 6 moderate, 1 severe). Of 111 who had a TI HST, 29% had a normal PSG and 71% had OSA (43 mild, 19 moderate, 17 severe). Individuals younger than 50 years old with a normal HST were more likely to have a normal PSG. Older age predicted diagnosis of OSA on PSG among individuals with a TI HST.

CONCLUSION

In this retrospective analysis of a clinical sample, when the HST is interpreted as normal in a younger patient population, the subsequent PSG is likewise normal in majority of the patients, although significant OSA is sometimes discovered. When a HST is read as TI, the majority of patients have OSA.

NuMoM2b Sleep-Disordered Breathing study: objectives and methods

OBJECTIVE

The objective of the Sleep Disordered Breathing substudy of the Nulliparous Pregnancy Outcomes Study Monitoring Mothers-to-be (nuMoM2b) is to determine whether sleep disordered breathing during pregnancy is a risk factor for adverse pregnancy outcomes.

STUDY DESIGN

NuMoM2b is a prospective cohort study of 10,037 nulliparous women with singleton gestations that was conducted across 8 sites with a central Data Coordinating and Analysis Center. The Sleep Disordered Breathing substudy recruited 3702 women from the cohort to undergo objective, overnight in-home assessments of sleep disordered breathing. A standardized level 3 home sleep test was performed between 6(0)-15(0) weeks' gestation (visit 1) and again between 22(0)-31(0) weeks' gestation (visit 3). Scoring of tests was conducted by a central Sleep Reading Center. Participants and their health care providers were notified if test results met "urgent referral" criteria that were based on threshold levels of apnea hypopnea indices, oxygen saturation levels, or electrocardiogram abnormalities but were not notified of test results otherwise. The primary pregnancy outcomes to be analyzed in relation to maternal sleep disordered breathing are preeclampsia, gestational hypertension, gestational diabetes mellitus, fetal growth restriction, and preterm birth.

RESULTS

Objective data were obtained at visit 1 on 3261 women, which was 88.1% of the studies that were attempted and at visit 3 on 2511 women, which was 87.6% of the studies that were attempted. Basic characteristics of the substudy cohort are reported in this methods article.

CONCLUSION

The substudy was designed to address important questions regarding the relationship of sleep-disordered breathing on the risk of preeclampsia and other outcomes of relevance to maternal and child health.

Effect of recording duration on the diagnostic accuracy of out-of-center sleep testing for obstructive sleep apnea

STUDY OBJECTIVES

This study investigated the minimum recording time needed during out-of-center sleep testing (OCST) to accurately diagnose the presence and severity of obstructive sleep apnea (OSA).

DESIGN AND SETTING

A retrospective analysis was conducted of OCSTs performed from October 2009 to May 2012 at the Mayo Clinic Center of Sleep Medicine using the portable Embletta™ system.

PATIENTS OR PARTICIPANTS

Demographic information was collected for patients who underwent OCSTs during the study period, including presenting symptoms, examination findings, and comorbidities.

INTERVENTION

Each study was divided into 60-, 120-, 180-, 240-, 300-, 360-, and 420-min intervals beginning at the recording start time to determine the respiratory event index (REI) for each of these time intervals. These interval values were then compared to the original REI derived from the total recording time (REITRT) by a paired t-test and concordance correlation coefficient (CCC).

MEASUREMENTS AND RESULTS

There were significant differences between the REITRT and the REI from the 60-min (P < 0.0001), 120-min (0.0001), 180-min (0.003) and 240-min (0.006) intervals with a lack of concordance, suggesting these intervals are poor diagnostic correlates for the REITRT. REIs determined at 300, 360, and 420 min were not significantly different from the REITRT and had highly significant CCCs, 0.963, 0.987, and 0.995, respectively.

CONCLUSIONS

The results suggest that at least 300 min recording time during out-of-center sleep testing is needed for accurate diagnosis of obstructive sleep apnea and determination of obstructive sleep apnea severity.

Detection of sleep disordered breathing and its central/obstructive character using nasal cannula and finger pulse oximeter

STUDY OBJECTIVE

To assess the accuracy of novel algorithms using an oximeter-based finger plethysmographic signal in combination with a nasal cannula for the detection and differentiation of central and obstructive apneas. The validity of single pulse oximetry to detect respiratory disturbance events was also studied.

METHODS

Patients recruited from four sleep laboratories underwent an ambulatory overnight cardiorespiratory polygraphy recording. The nasal flow and photoplethysmographic signals of the recording were analyzed by automated algorithms. The apnea hypopnea index (AHI(auto)) was calculated using both signals, and a respiratory disturbance index (RDI(auto)) was calculated from photoplethysmography alone. Apnea events were classified into obstructive and central types using the oximeter derived pulse wave signal and compared with manual scoring.

RESULTS

Sixty-six subjects (42 males, age 54 ± 14 yrs, body mass index 28.5 ± 5.9 kg/m(2)) were included in the analysis. AHI(manual) (19.4 ± 18.5 events/h) correlated highly significantly with AHI(auto) (19.9 ± 16.5 events/h) and RDI(auto) (20.4 ± 17.2 events/h); the correlation coefficients were r = 0.94 and 0.95, respectively (p < 0.001) with a mean difference of -0.5 ± 6.6 and -1.0 ± 6.1 events/h. The automatic analysis of AHI(auto) and RDI(auto) detected sleep apnea (cutoff AHI(manual) ≥ 15 events/h) with a sensitivity/specificity of 0.90/0.97 and 0.86/0.94, respectively. The automated obstructive/central apnea indices correlated closely with manually scoring (r = 0.87 and 0.95, p < 0.001) with mean difference of -4.3 ± 7.9 and 0.3 ± 1.5 events/h, respectively.

CONCLUSIONS

Automatic analysis based on routine pulse oximetry alone may be used to detect sleep disordered breathing with accuracy. In addition, the combination of photoplethysmographic signals with a nasal flow signal provides an accurate distinction between obstructive and central apneic events during sleep.