Automated remote sleep monitoring needs uncertainty quantification.

Wearable electroencephalography devices emerge as a cost-effective and ergonomic alternative to gold-standard polysomnography, paving the way for better health monitoring and sleep disorder screening. Machine learning allows to automate sleep stage classification, but trust and reliability issues have hampered its adoption in clinical applications. Estimating uncertainty is a crucial factor in enhancing reliability by identifying regions of heightened and diminished confidence. In this study, we used an uncertainty-centred machine learning pipeline, U-PASS, to automate sleep staging in a challenging real-world dataset of single-channel electroencephalography and accelerometry collected with a wearable device from an elderly population. We were able to effectively limit the uncertainty of our machine learning model and to reliably inform clinical experts of which predictions were uncertain to improve the machine learning model's reliability. This increased the five-stage sleep-scoring accuracy of a state-of-the-art machine learning model from 63.9% to 71.2% on our dataset. Remarkably, the machine learning approach outperformed the human expert in interpreting these wearable data. Manual review by sleep specialists, without specific training for sleep staging on wearable electroencephalography, proved ineffective. The clinical utility of this automated remote monitoring system was also demonstrated, establishing a strong correlation between the predicted sleep parameters and the reference polysomnography parameters, and reproducing known correlations with the apnea-hypopnea index. In essence, this work presents a promising avenue to revolutionize remote patient care through the power of machine learning by the use of an automated data-processing pipeline enhanced with uncertainty estimation.

Acoustic Stimulation to Improve Slow-Wave Sleep in Alzheimer's Disease: A Multiple Night At-Home Intervention.

OBJECTIVES: To investigate the efficacy of closed-loop acoustic stimulation (CLAS) during slow-wave sleep (SWS) to enhance slow-wave activity (SWA) and SWS in patients with Alzheimer's disease (AD) across multiple nights and to explore associations between stimulation, participant characteristics, and individuals' SWS response. DESIGN: A 2-week, open-label at-home intervention study utilizing the DREEM2 headband to record sleep data and administer CLAS during SWS. SETTING AND PARTICIPANTS: Fifteen older patients with AD (6 women, mean age: 76.27 [SD = 6.06], mean MOCA-score: 16.07 [SD = 6.94]), living at home with their partner, completed the trial. INTERVENTION: Patients first wore the device for two baseline nights, followed by 14 nights during which the device was programmed to randomly either deliver acoustic stimulations of 50 ms pink noise (± 40 dB) targeted to the slow-wave up-phase during SWS or only mark the wave (sham). RESULTS: On a group level, stimulation significantly enhanced SWA and SWS with consistent SWS enhancement throughout the intervention. However, substantial variability existed in individual responses to stimulation. Individuals received more stimulations on nights with increased SWS compared to baseline than on nights with no change or a decrease. In individuals, having lower baseline SWS correlated with receiving fewer stimulations on average during the intervention. CONCLUSION: CLAS during SWS is a promising nonpharmacological method to enhance SWA and SWS in AD. However, patients with lower baseline SWS received fewer stimulations during the intervention, possibly resulting in less SWS enhancement. Individual variability in response to stimulation underscores the need to address personalized stimulation parameters in future research and therapy development.

In search of a cut-off apnea-hypopnea index in type 3 home portable monitors to diagnose and treat obstructive sleep apnea: a mathematical simulation.

The American Academy of Sleep Medicine (AASM) uses similar apnea-hypopnea index (AHI) cut-off values to diagnose and define severity of sleep apnea independent of the technique used: in-hospital polysomnography (PSG) or type 3 portable monitoring (PM). Taking into account that PM theoretically might underestimate the AHI, we explored whether a lower cut-off would be more appropriate. We performed mathematical re-calculations on the diagnostic PSG-AHI (scored using AASM 1999 rules) of 865 consecutive patients with an AHI of ≥20 events/h who started continuous positive airway pressure (CPAP). For a PSG-AHI of ≥15 events/h re-scored using AASM 2012 rules (PSG-AHIAASM2012 ), a PM-respiratory event index (REI)AASM2012 cut-off point of ≥15 events/h resulted in a post-test probability of 100% of having the disease, but with negative tests in 57.1%. A PM-REIAASM2012 cut-off of 8 events/h, still resulted in a positive post-test probability of 100% but with negative tests in only 34.3%. Combination of the cut-off values with clinical estimation of being 'at high risk' based on Epworth Sleepiness Scale (ESS) and Berlin Questionnaire scores only resulted in a small reduction in the percentage of negative tests (respectively 52.7% and 32.7%). After 6 months, CPAP adherence was not lower using the PM-REIAASM 2012 cut-off ≥8 events/h in comparison to ≥15 events/h (median 5.7 vs. 5.8 h/night, p = 0.368) and the reduction in ESS was similar too (median -4 and -5 points, p = 0.083). Consequently, using a lower PM-REIAASM2012 cut-off could result in cost savings because of less negative studies and lesser need for a confirmatory PSG or a performance of a CPAP trial.

Sleep-disordered breathing after lung transplantation: An observational cohort study.

Data concerning sleep-disordered breathing (SDB) after lung transplantation (LTX) are scarce. This study aims to analyze prevalence, associated factors, and impact on survival of moderate to severe SDB in a large cohort of consecutive LTX patients (n = 219). Patients underwent a diagnostic polysomnography 1 year after LTX. Moderate to severe SDB was present in 57.5% of patients, with the highest prevalence in chronic obstructive pulmonary disease/emphysema (71.1%) and pulmonary fibrosis (65.1%). SDB patients were older, mostly male, and had higher body mass index and neck circumference. Nocturnal diastolic and 24-hour blood pressures were higher in SDB patients. In 45 patients, polysomnography was also performed pre-LTX. Compared to pre-LTX, mean apnea/hypopnea index (AHI) increased significantly after LTX. A significant correlation was seen between lung function parameters and AHI, suggesting a role of decreased caudal traction on the pharynx. Presence of SDB had no impact on mortality or prevalence of chronic lung allograft dysfunction. However, survival was better in continuous positive airway pressure (CPAP) compliant SDB patients compared to SDB patients without CPAP treatment. These findings may be pertinent for systematic screening of SDB after LTX.

Obstructive sleep apnea in head and neck cancer survivors.

IMPORTANCE: Chronic fatigue is present in 33.0% of all head and neck cancer (HNC) survivors; this impacts their quality of life negatively. A plausible cause is obstructive sleep apnea (OSA) after HNC treatment. However, studies regarding this topic are scarce. OBJECTIVE: To confirm if OSA is more prevalent after receiving radiotherapy for HNC. In addition, investigation of the risk factors for developing OSA in this population. DESIGN: A retrospective review of prospective data. METHODS: Treatment for HNC took place between 2016 and 2017 at the University Hospital of Leuven. One hundred sixty-four patients were eligible for participating in this study. Sixty-five responded (39.4%). Upon consulting their medical files, 15 patients were excluded based on the in- and exclusion criteria. Presence of OSA was estimated using standardized questionnaires, namely the Berlin Questionnaire, the Epworth Sleepiness Scale, and the CIS-20. This was compared to the proportion of OSA in the general population. RESULTS: Fifty patients (33 men, 17 women) with a mean age of 64.2 years (range 32-88) were included. Based on the questionnaires, OSA was suspected in twenty. The prevalence of suspected OSA in our study group (40.0%) was significantly greater (p < 0.0001) than our estimated prevalence of OSA in the general population (10.9%). No significant risk factors could be identified. CONCLUSION: Patients treated for HNC are at risk of developing OSA. When complaints of fatigue and sleeping problems persist, referral to a sleep clinic is suggested. Further investigation remains necessary to identify potential risk factors along with prevention and treatment strategies.

Capacitively-Coupled ECG and Respiration for Sleep-Wake Prediction and Risk Detection in Sleep Apnea Patients.

Obstructive sleep apnea (OSA) patients would strongly benefit from comfortable home diagnosis, during which detection of wakefulness is essential. Therefore, capacitively-coupled electrocardiogram (ccECG) and bioimpedance (ccBioZ) sensors were used to record the sleep of suspected OSA patients, in parallel with polysomnography (PSG). The three objectives were quality assessment of the unobtrusive signals during sleep, prediction of sleep-wake using ccECG and ccBioZ, and detection of high-risk OSA patients. First, signal quality indicators (SQIs) determined the data coverage of ccECG and ccBioZ. Then, a multimodal convolutional neural network (CNN) for sleep-wake prediction was tested on these preprocessed ccECG and ccBioZ data. Finally, two indices derived from this prediction detected patients at risk. The data included 187 PSG recordings of suspected OSA patients, 36 (dataset "Test") of which were recorded simultaneously with PSG, ccECG, and ccBioZ. As a result, two improvements were made compared to prior studies. First, the ccBioZ signal coverage increased significantly due to adaptation of the acquisition system. Secondly, the utility of the sleep-wake classifier increased as it became a unimodal network only requiring respiratory input. This was achieved by using data augmentation during training. Sleep-wake prediction on "Test" using PSG respiration resulted in a Cohen's kappa (κ) of 0.39 and using ccBioZ in κ = 0.23. The OSA risk model identified severe OSA patients with a κ of 0.61 for PSG respiration and κ of 0.39 using ccBioZ (accuracy of 80.6% and 69.4%, respectively). This study is one of the first to perform sleep-wake staging on capacitively-coupled respiratory signals in suspected OSA patients and to detect high risk OSA patients based on ccBioZ. The technology and the proposed framework could be applied in multi-night follow-up of OSA patients.

Supervised SVM Transfer Learning for Modality-Specific Artefact Detection in ECG.

The electrocardiogram (ECG) is an important diagnostic tool for identifying cardiac problems. Nowadays, new ways to record ECG signals outside of the hospital are being investigated. A promising technique is capacitively coupled ECG (ccECG), which allows ECG signals to be recorded through insulating materials. However, as the ECG is no longer recorded in a controlled environment, this inevitably implies the presence of more artefacts. Artefact detection algorithms are used to detect and remove these. Typically, the training of a new algorithm requires a lot of ground truth data, which is costly to obtain. As many labelled contact ECG datasets exist, we could avoid the use of labelling new ccECG signals by making use of previous knowledge. Transfer learning can be used for this purpose. Here, we applied transfer learning to optimise the performance of an artefact detection model, trained on contact ECG, towards ccECG. We used ECG recordings from three different datasets, recorded with three recording devices. We showed that the accuracy of a contact-ECG classifier improved between 5 and 8% by means of transfer learning when tested on a ccECG dataset. Furthermore, we showed that only 20 segments of the ccECG dataset are sufficient to significantly increase the accuracy.

Benchmarking Transfer Entropy Methods for the Study of Linear and Nonlinear Cardio-Respiratory Interactions.

Transfer entropy (TE) has been used to identify and quantify interactions between physiological systems. Different methods exist to estimate TE, but there is no consensus about which one performs best in specific applications. In this study, five methods (linear, k-nearest neighbors, fixed-binning with ranking, kernel density estimation and adaptive partitioning) were compared. The comparison was made on three simulation models (linear, nonlinear and linear + nonlinear dynamics). From the simulations, it was found that the best method to quantify the different interactions was adaptive partitioning. This method was then applied on data from a polysomnography study, specifically on the ECG and the respiratory signals (nasal airflow and respiratory effort around the thorax). The hypothesis that the linear and nonlinear components of cardio-respiratory interactions during light and deep sleep change with the sleep stage, was tested. Significant differences, after performing surrogate analysis, indicate an increased TE during deep sleep. However, these differences were found to be dependent on the type of respiratory signal and sampling frequency. These results highlight the importance of selecting the appropriate signals, estimation method and surrogate analysis for the study of linear and nonlinear cardio-respiratory interactions.

Evaluation of a Commercial Ballistocardiography Sensor for Sleep Apnea Screening and Sleep Monitoring.

There exists a technological momentum towards the development of unobtrusive, simple, and reliable systems for long-term sleep monitoring. An off-the-shelf commercial pressure sensor meeting these requirements is the Emfit QS. First, the potential for sleep apnea screening was investigated by revealing clusters of contaminated and clean segments. A relationship between the irregularity of the data and the sleep apnea severity class was observed, which was valuable for screening (sensitivity 0.72, specificity 0.70), although the linear relation was limited ( R 2 of 0.16). Secondly, the study explored the suitability of this commercial sensor to be merged with gold standard polysomnography data for future sleep monitoring. As polysomnography (PSG) and Emfit signals originate from different types of sensor modalities, they cannot be regarded as strictly coupled. Therefore, an automated synchronization procedure based on artefact patterns was developed. Additionally, the optimal position of the Emfit for capturing respiratory and cardiac information similar to the PSG was identified, resulting in a position as close as possible to the thorax. The proposed approach demonstrated the potential for unobtrusive screening of sleep apnea patients at home. Furthermore, the synchronization framework enabled supervised analysis of the commercial Emfit sensor for future sleep monitoring, which can be extended to other multi-modal systems that record movements during sleep.

Definition, discrimination, diagnosis and treatment of central breathing disturbances during sleep.

The complexity of central breathing disturbances during sleep has become increasingly obvious. They present as central sleep apnoeas (CSAs) and hypopnoeas, periodic breathing with apnoeas, or irregular breathing in patients with cardiovascular, other internal or neurological disorders, and can emerge under positive airway pressure treatment or opioid use, or at high altitude. As yet, there is insufficient knowledge on the clinical features, pathophysiological background and consecutive algorithms for stepped-care treatment. Most recently, it has been discussed intensively if CSA in heart failure is a "marker" of disease severity or a "mediator" of disease progression, and if and which type of positive airway pressure therapy is indicated. In addition, disturbances of respiratory drive or the translation of central impulses may result in hypoventilation, associated with cerebral or neuromuscular diseases, or severe diseases of lung or thorax. These statements report the results of an European Respiratory Society Task Force addressing actual diagnostic and therapeutic standards. The statements are based on a systematic review of the literature and a systematic two-step decision process. Although the Task Force does not make recommendations, it describes its current practice of treatment of CSA in heart failure and hypoventilation.

Feasibility of diaphragm pacing in patients after bilateral lung transplantation.

Recent animal studies and intraoperative studies in humans suggested that phrenic nerve stimulation could attenuate ventilator-induced diaphragm dysfunction. The purpose of the present study is to examine the safety and feasibility of diaphragm pacing during the weaning process after bilateral lung transplantation. Four patients, suffering from chronic pulmonary disease, were included, and diaphragm pacing was evaluated after lung transplantation. Implantation of electrodes at the end of the lung transplant procedure was possible in three of the four patients. In all implanted patients, stimulation of the diaphragm could trigger the ventilator. Implanted electrodes were completely removed by percutaneous retraction after up to 7 days of pacing. Adverse events related to pacing included occurrence of pain. Diaphragm pacing with temporary electrodes, inserted during surgery, is feasible and is able to trigger the ventilator in patients after bilateral lung transplantation. The use of intradiaphragmatic electrodes creates the additional opportunity to monitor the evolution of diaphragm electromyography during the postoperative weaning process.

Randomized cross-over trial of ventilator modes during non-invasive ventilation titration in amyotrophic lateral sclerosis.

BACKGROUND AND OBJECTIVE: Non-invasive ventilation (NIV) improves survival, quality of life and sleep in patients with amyotrophic lateral sclerosis (ALS). Nevertheless, NIV titration is conducted in different ways. We aim to provide more insight into NIV titration by comparing the effects of a spontaneous (S) and spontaneous-timed (ST) modes on gas exchange, sleep architecture and patient-ventilator asynchronies (PVAs). METHODS: After an initial night of NIV titration, patients were randomized to S or ST mode in a cross-over design. NIV was titrated using polysomnography, oximetry (oxygen saturation, SpO2 %) and transcutaneous carbon dioxide (PtcCO2 ) measurement. PVAs were analysed breath-by-breath. RESULTS: Thirteen patients were analysed after inclusion. ST mode showed better results in gas exchange (minimal SpO2 %: 83 (80-89)% vs 87 (84-89)%; oxygen desaturation index: 15 (5-28)/h sleep vs 7 (3-9)/h sleep; PtcCO2 >55 mm Hg: 20 (0-59)% vs 0 (0-27)% total sleep time for S and ST mode, respectively, all P < 0.05) and respiratory events (obstructive: 8.9 (1.2-18.3)/h sleep vs 1.8 (0.3-4.9)/h sleep and central: 2.6 (0.4-14.1)/h sleep vs 0.2 (0.0-1.1)/h sleep for S and ST mode, respectively, both P < 0.01). No differences in sleep architecture were found. Ineffective efforts and respiratory events were more frequently present in S mode. Nevertheless, four patients were discharged on S mode as these patients showed clinically better results for sleep architecture and PVA during the night on S mode. CONCLUSION: ST mode shows better results in gas exchange, respiratory events and PVA. Nevertheless, accurate NIV titration remains necessary as some patients show equal or better results when using the S mode.

Possible influence of AMPD1 on cholinergic neurotransmission and sleep.

It is known that adenosine excess due to monophosphate deaminase deficiency (AMPD1) can be linked to muscle problems. Recently, Perumal et al., 2014 reported a first case of possible impact of AMPD1 on sleep, REM sleep and cholinergic neurotransmission. We report a second patient with similar sleep complaints: long sleep duration with residual daytime sleepiness and a need to sleep after exercise. On polysomnography we observed a long sleep duration, with high sleep efficiency and a SOREMP; on MSLT a shortened sleep latency and 4 SOREMPS were observed. Frequency power spectral heart rate analysis during slow wave sleep, REM sleep and wakefulness revealed an increased parasympathetic tone. In conclusion, AMPD1 could have a profound influence on cholinergic neurotransmission and sleep; further studies are mandatory.

Central sleep apnea: emphasizing recognition and differentiation.

INTRODUCTION: Central sleep apnea (CSA) is a sleep-related breathing disorder in which the effort to breathe is intermittently diminished or absent. CSA is a common disorder among patients with different cardiovascular disorders, including heart failure. In addition, a growing number of medications have been shown to induce CSA and CSA can emerge after initiation of treatment for obstructive sleep apnea. Accumulating evidence shows that CSA is a heterogeneous disorder with individual differences in clinical and biological characteristics and/or underlying pathophysiological mechanisms. AREAS COVERED: This narrative review offers an overview of the diagnostic aspects and classification of CSA, with an emphasis on heart failure patients, patients with CSA due to a medication and treatment-emergent CSA. The importance of evaluation of prognostic biomarkers in patients with different types of CSA is discussed. This narrative review synthesizes literature on CSA sourced from the PubMed database up to February 2024. EXPERT OPINION: CSA presents a remarkably diverse disorder, with treatment modalities exhibiting potentially varied efficacy across its various phenotypes. This highlights the imperative for tailored management strategies that are rooted in phenotype classification.

U-PASS: An uncertainty-guided deep learning pipeline for automated sleep staging.

With the increasing prevalence of machine learning in critical fields like healthcare, ensuring the safety and reliability of these systems is crucial. Estimating uncertainty plays a vital role in enhancing reliability by identifying areas of high and low confidence and reducing the risk of errors. This study introduces U-PASS, a specialized human-centered machine learning pipeline tailored for clinical applications, which effectively communicates uncertainty to clinical experts and collaborates with them to improve predictions. U-PASS incorporates uncertainty estimation at every stage of the process, including data acquisition, training, and model deployment. Training is divided into a supervised pre-training step and a semi-supervised recording-wise finetuning step. We apply U-PASS to the challenging task of sleep staging and demonstrate that it systematically improves performance at every stage. By optimizing the training dataset, actively seeking feedback from domain experts for informative samples, and deferring the most uncertain samples to experts, U-PASS achieves an impressive expert-level accuracy of 85% on a challenging clinical dataset of elderly sleep apnea patients. This represents a significant improvement over the starting point at 75% accuracy. The largest improvement gain is due to the deferral of uncertain epochs to a sleep expert. U-PASS presents a promising AI approach to incorporating uncertainty estimation in machine learning pipelines, improving their reliability and unlocking their potential in clinical settings.

Catathrenia in severe obstructive sleep apnea: A novel entity never described before.

STUDY OBJECTIVES: Catathrenia, derived from the Greek κατά (kata) meaning below and θρηνώ (threnia) to lament, is characterized by expiratory groaning episodes during sleep. In a case series of nine patients with severe obstructive sleep apnea, we observed a peculiar groaning entity that has not been described before. METHODS: We described and illustrated the cases with polysomnographic tracings and additional audio recordings. RESULTS: All patients were men, obese (body mass index 39 ± 6 kg/m2) with an apnea-hypopnea index ranging from 47 to 125/h. In addition, we identified groaning events that were consistently preceded by a cortical arousal associated with a "rescue" respiration after an obstructive hypopnea or apnea. These events exhibited characteristics of "mixed apnea's", but the "central apnea-like part" was a prolonged expiratory groaning phase, with immediately after the terminal expiratory snort appearance of an obstructive apnea. In case the duration of this expiration was at least 10 s we calculated these events separately and the index was 8.4 ± 7.7/h. More rarely (index 0.6 ± 0.5/h) a "central apnea mimicking event" with groaning not followed by an obstruction, was observed. We also observed groaning episodes during expiration with a shorter duration (less than 10 s), not calculated separately. Positive airway pressure, which was well tolerated, eliminated these events. CONCLUSIONS: This novel catathrenia entity preceded by a cortical arousal and "rescue" respiration in response to obstructive events is intriguing. Possible explanations for these observations are further discussed in this article.

Insights into the Use of Point-of-Care Ultrasound for Diagnosing Obstructive Sleep Apnea.

Obstructive sleep apnea (OSA) is a sleeping disorder caused by complete or partial disturbance of breathing during the night. Existing screening methods include questionnaire-based evaluations which are time-consuming, vary in specificity, and are not globally adopted. Point-of-care ultrasound (PoCUS), on the other hand, is a painless, inexpensive, portable, and useful tool that has already been introduced for the evaluation of upper airways by anesthetists. PoCUS could also serve as a potential screening tool for the diagnosis of OSA by measuring different airway parameters, including retropalatal pharynx transverse diameter, tongue base thickness, distance between lingual arteries, lateral parapharyngeal wall thickness, palatine tonsil volume, and some non-airway parameters like carotid intima-media thickness, mesenteric fat thickness, and diaphragm characteristics. This study reviewed previously reported studies to highlight the importance of PoCUS as a potential screening tool for OSA.

Recovery from shift work.

One fifth of today's workforce is engaged in shift work and exposed to various mental and physical health risks including shift work disorder. Efficiently recovering from shift work through physical and mental interventions allows us to mitigate negative effects on health, enables a better work-life balance and enhances our overall wellbeing. The aim of this review is to provide a state-of-the-art overview of the available literature. The role of sleep timing and naps, light therapy and psychotherapy, diet and exercise in recovery from shift work is presented here. We further review the impact of shift schedules and social support on post-shift unwinding.

Short-term positive effects of a mandibular advancement device in a selected phenotype of patients with moderate obstructive sleep apnea: a prospective study.

STUDY OBJECTIVES: To evaluate (determinants of) treatment success of mandibular advancement device application in a selected phenotype of patients with obstructive sleep apnea (OSA). METHODS: Ninety nonobese patients with moderate OSA (obstructive apnea-hypopnea index [OAHI] ≥ 15 and < 30 events/h) without comorbidities were prospectively included. Polysomnography was performed at baseline and with a mandibular advancement device. A drug-induced sleep endoscopy with jaw thrust was performed in 83%. RESULTS: OAHI reduction ≥ 50% was observed in 73%, OAHI reduction ≥ 50% with OAHI < 10 events/h in 70%, and complete OSA resolution (OAHI < 5 events/h) in 40%. Patients with nonpositional OSA showed a significantly higher rate of complete OSA resolution: Posttest probability increased to 67%. In patients with total disappearance of collapse at velum level and at all levels during drug-induced sleep endoscopy with jaw thrust, the drop in OAHI was impressive with an infinitively high positive likelihood ratio. However, the proportion of patients having nonpositional OSA or the drug-induced sleep endoscopy characteristics as described above was < 20%. The change in snoring disturbance based on a visual analog scale was 76% (interquartile range 40-89%, P < .001) and a statistically significant amelioration in Epworth Sleepiness Scale (especially in somnolent subjects) was observed. High adherence was reported. CONCLUSIONS: In this predefined OSA phenotype, a mandibular advancement device was effective in reduction of OAHI and in amelioration of symptoms. Stratification by nonpositional OSA and findings on drug-induced sleep endoscopy with jaw thrust increased treatment success defined as reduction in OAHI. However, the clinical relevance can be questioned because only a small number of patients demonstrated these characteristics. CITATION: Buyse B, Nguyen PAH, Leemans J, et al. Short-term positive effects of a mandibular advancement device in a selected phenotype of patients with moderate obstructive sleep apnea: a prospective study. J Clin Sleep Med. 2023;19(1):5-16.

Obstructive Sleep Apnea Screening by Joint Saturation Signal Analysis and PPG-derived Pulse Rate Oscillations.

Obstructive sleep apnea (OSA) is a high-prevalence disease in the general population, often underdiagnosed. The gold standard in clinical practice for its diagnosis and severity assessment is the polysomnography, although in-home approaches have been proposed in recent years to overcome its limitations. Today's ubiquitously presence of wearables may become a powerful screening tool in the general population and pulse-oximetry-based techniques could be used for early OSA diagnosis. In this work, the peripheral oxygen saturation together with the pulse-to-pulse interval (PPI) series derived from photoplethysmography (PPG) are used as inputs for OSA diagnosis. Different models are trained to classify between normal and abnormal breathing segments (binary decision), and between normal, apneic and hypopneic segments (multiclass decision). The models obtained 86.27% and 73.07% accuracy for the binary and multiclass segment classification, respectively. A novel index, the cyclic variation of the heart rate index (CVHRI), derived from PPI's spectrum, is computed on the segments containing disturbed breathing, representing the frequency of the events. CVHRI showed strong Pearson's correlation (r) with the apnea-hypopnea index (AHI) both after binary (r=0.94, p 0.001) and multiclass (r=0.91, p 0.001) segment classification. In addition, CVHRI has been used to stratify subjects with AHI higher/lower than a threshold of 5 and 15, resulting in 77.27% and 79.55% accuracy, respectively. In conclusion, patient stratification based on the combination of oxygen saturation and PPI analysis, with the addition of CVHRI, is a suitable, wearable friendly and low-cost tool for OSA screening at home.