Apnoea and hypopnoea scoring for people with spinal cord injury: new thresholds for sleep disordered breathing diagnosis and severity classification.

STUDY DESIGN: Descriptive study. OBJECTIVES: To determine the effect of respiratory event rule-set changes on the apnoea hypopnoea index, and diagnostic and severity thresholds in people with acute and chronic spinal cord injury. SETTING: Eleven acute spinal cord injury inpatient hospitals across Australia, New Zealand, Canada and England; community dwelling chronic spinal cord injury patients in their own homes. METHODS: Polysomnography of people with acute (n = 24) and chronic (n = 78) tetraplegia were reanalysed from 1999 American Academy of Sleep Medicine (AASM) respiratory scoring, to 2007 AASM 'alternative' and 2012 AASM respectively. Equivalent cut points for published 1999 AASM sleep disordered breathing severity ranges were calculated using receiver operator curves, and results presented alongside analyses from the able-bodied. RESULTS: In people with tetraplegia, shift from 1999 AASM to 2007 AASM 'alternative' resulted in a 22% lower apnoea hypopnoea index, and to 2012 AASM a 17% lower index. In people with tetraplegia, equivalent cut-points for 1999 AASM severities of 5,15 and 30 were calculated at 2.4, 8.1 and 16.3 for 2007 AASM 'alternative' and 3.2, 10.0 and 21.2 for 2012 AASM. CONCLUSION: Interpreting research, prevalence and clinical polysomnography results conducted over different periods requires knowledge of the relationship between different rule-sets, and appropriate thresholds for diagnosis of disease. SPONSORSHIP: This project was proudly supported by the Traffic Accident Commission (Program grant) and the National Health and Medical Research Council (PhD stipend 616605).

A Deep Learning Architecture for Temporal Sleep Stage Classification Using Multivariate and Multimodal Time Series.

Sleep stage classification constitutes an important preliminary exam in the diagnosis of sleep disorders. It is traditionally performed by a sleep expert who assigns to each 30 s of the signal of a sleep stage, based on the visual inspection of signals such as electroencephalograms (EEGs), electrooculograms (EOGs), electrocardiograms, and electromyograms (EMGs). We introduce here the first deep learning approach for sleep stage classification that learns end-to-end without computing spectrograms or extracting handcrafted features, that exploits all multivariate and multimodal polysomnography (PSG) signals (EEG, EMG, and EOG), and that can exploit the temporal context of each 30-s window of data. For each modality, the first layer learns linear spatial filters that exploit the array of sensors to increase the signal-to-noise ratio, and the last layer feeds the learnt representation to a softmax classifier. Our model is compared to alternative automatic approaches based on convolutional networks or decisions trees. Results obtained on 61 publicly available PSG records with up to 20 EEG channels demonstrate that our network architecture yields the state-of-the-art performance. Our study reveals a number of insights on the spatiotemporal distribution of the signal of interest: a good tradeoff for optimal classification performance measured with balanced accuracy is to use 6 EEG with 2 EOG (left and right) and 3 EMG chin channels. Also exploiting 1 min of data before and after each data segment offers the strongest improvement when a limited number of channels are available. As sleep experts, our system exploits the multivariate and multimodal nature of PSG signals in order to deliver the state-of-the-art classification performance with a small computational cost.

Accuracy of residual apnea-hypopnea index obtained using the continuous positive airway pressure device: application of new version 2.0 scoring rules for respiratory events during sleep.

PURPOSE: Continuous positive airway pressure (CPAP) devices can estimate apnea-hypopnea index (AHI) using respiratory event detection algorithms. In 2012, rules for manually scoring respiratory events during sleep were updated to version 2.0. The purpose of the present study was to compare residual AHI determined using the Sleepstyle HC608 CPAP device (HC) with those determined by the new manual scoring (NM) rules during CPAP titration in patients with obstructive sleep apnea (OSA). METHODS: Fifty-seven patients underwent CPAP titration with HC. Correlations were assessed between AHI determined by NM and HC. The AHI, the apnea index (AI), and the hypopnea index (HI) were evaluated separately. RESULTS: The mean AHI as assessed using diagnostic polysomnography (PSG) was 53.9 ± 22.4. During CPAP titration, respiratory events were effectively suppressed (HC-AHI, 4.2 ± 6.0; NM-AHI, 6.0 ± 5.8). Lower HI and AHI were obtained using HC compared to NM (HC-HI, 2.9 ± 3.6 and NM-HI, 5.2 ± 4.2, p < 0.001; HC-AHI, 4.2 ± 6.0 and NM-AHI, 6.0 ± 5.8, p < 0.001). Additionally, HC reported higher AI compared to NM (HC-AI, 1.3 ± 2.8; NM-AI, 0.9 ± 2.2, p = 0.002). NM-AI (ß = 1.017, p < 0.001), NM-HI (ß = -0.599, p < 0.001), and NM-arousal index (ß = -0.058, p = 0.042) were associated with greater differences between HC-AHI and NM-AHI in multivariate regression analysis. CONCLUSIONS: Our findings indicate differences in scoring respiratory events between our CPAP device and new version 2.0 manual scoring and suggest that residual AHI values should be carefully interpreted.

Process and outcome for international reliability in sleep scoring.

OBJECTIVES: The aim was to evaluate the inter-rater reliability in scoring sleep stages in two sleep labs in Berlin Germany and Beijing China. METHODS: The subjects consist of polysomnography (PSGs) from 15 subjects in a German sleep laboratory, with 7 mild to moderate sleep apnea hypopnea syndrome (SAHS) patients and 8 healthy controls, and PSGs from 15 narcolepsy patients in a Chinese sleep laboratory. Five experienced technologists including two Chinese and three Germans without common training scored the PSGs following the 2007 AASM manual except the EEG signals included only two EEG leads (C3/A2 and C4/A1). Differences in inter-scorer agreement were analyzed based on epoch-by-epoch comparison by means of Cohen's κ, and quantitative sleep parameters by means of intra-class correlation coefficients. RESULTS: Inter-laboratory epoch-by-epoch agreement comparison between scorers from the two countries yielded a moderate agreement with a mean κ value of 0.57 for controls, 0.58 for SAHS, and 0.54 for narcolepsy. When compared with controls, the inter-scoring agreement is higher for wake and N3 stage scoring in SAHS and N1 and N3 scoring in narcolepsy (p < 0.05). The only sleep stage with lower scoring agreement in both SAHS (κ 0.69 vs. 0.79, p = 0.034) and narcolepsy (0.66 vs 0.79, p = 0.022) was stage REM. Inter-laboratory comparisons showed that the most common combinations of deviating scorings were N1 and N2, N2 and N3, and N1 and wake. A 6.5 % deviating scoring rate of wake and REM and a 13.4 % deviating scoring rate of N1 and REM indicated that inter-laboratory scoring in narcolepsy was about twice as in SAHS and controls confused. This was further confirmed by agreement analysis of quantitative parameters using intra-class correlation coefficients ICC(2,1) indicating REM sleep scoring agreement was lower in narcolepsy than in controls (p < 0.05). CONCLUSION: Low REM stage scoring agreement exists for narcoleptics and SAHS, indicating the necessity to study sleep stage scoring agreement for a specific sleep disorder. Intensive training is needed for the scoring of sleep in international multiple center studies to improve the scoring agreement.

Implications of revised AASM rules on scoring apneic and hypopneic respiratory events in patients with heart failure with nocturnal Cheyne-Stokes respiration.

STUDY OBJECTIVES: This study investigated the implications of the revised scoring rules of the American Academy of Sleep Medicine (AASM) in patients with heart failure (HF) with Cheyne-Stokes respiration (CSR). METHODS: Ninety-one patients (NYHA ≥II, LVEF ≤45 %; age 73.6 ± 11.3 years old; 81 male subjects) with documented CSR underwent 8 h of cardiorespiratory polygraphy recordings. Those were analyzed by a single scorer strictly applying the 2007 recommended, 2007 alternative, and the 2012 scoring rules. RESULTS: Compared with the AASM 2007 recommended rules, apnea-hypopnea index (AHI) and hypopnea index (HI) increased significantly when the 2007 alternative and 2012 rules were applied (AHI 34.1 ± 13.5/h vs 37.6 ± 13.2/h vs 38.3 ± 13.2/h, respectively; HI 10.2 ± 9.4/h vs 13.7 ± 10.7/h vs 14.4 ± 11.0/h, respectively; all p < 0.001). Duration of CSR increased significantly with the alternate versus recommended 2007 rules (182.2 ± 117.0 vs 170.1 ± 115.0 min; p ≤ 0.001); there was a significant decrease in CSR duration for the 2012 versus 2007 alternative rules (182.2 ± 117.0 vs 166.7 ± 115.4 min; p ≤ 0.001). CONCLUSION: AHI was higher using the AASM 2012 scoring rules due to a less strict definition of hypopnea. Data on the prognostic effects of CSR in patients with HF and the benefits of treatment are mostly based on the AASM 2007 recommended rules, so differences between these and the newer version need to be taken into account.

Automatic sleep staging from ventilator signals in non-invasive ventilation.

Non-invasive ventilation (NIV), a recognized treatment for chronic hypercapnic respiratory failure, is predominantly applied at night. Nevertheless, the quality of sleep is rarely evaluated due to the required technological complexity. A new technique for automatic sleep staging is here proposed for patients treated by NIV. This new technique only requires signals (airflow and hemoglobin oxygen saturation) available in domiciliary ventilators plus a photo-plethysmogram, a signal already managed by some ventilators. Consequently, electroencephalogram, electrooculogram, electromyogram, and electrocardiogram recordings are not needed. Cardiorespiratory features are extracted from the three selected signals and used as input to a Support Vector Machine (SVM) multi-class classifier. Two different types of sleep scoring were investigated: the first type was used to distinguish three stages (wake, REM sleep and nonREM sleep), and the second type was used to evaluate five stages (wake, REM sleep, N1, N2 and N3 stages). Patient-dependent and patient-independent classifiers were tested comparing the resulting hypnograms with those obtained from visual/manual scoring by a sleep specialist. An average accuracy of 91% (84%) was obtained with three-stage (five-stage) patient-dependent classifiers. With patient-independent classifiers, an average accuracy of 78% (62%) was obtained when three (five) sleep stages were scored. Also if the PPG-based and flow features are left out, a reduction of 4.5% (resp. 5%) in accuracy is observed for the three-stage (resp. five-stage) cases. Our results suggest that long-term sleep evaluation and nocturnal monitoring at home is feasible in patients treated by NIV. Our technique could even be integrated into ventilators.

Coding and billing for home (out-of-center) sleep testing.

Obstructive sleep apnea is increasingly recognized as a comorbidity in many medical illnesses. This has resulted in an increasing need for sleep testing, which is not entirely met by the currently available sleep laboratory facilities. Home, or out-of-center, sleep testing is an alternative to in-laboratory studies. However, coding and billing for home studies is not as straightforward as it is for in-laboratory studies. This article reviews the process of coding and billing for sleep studies done in an unattended setting.

Computer-assisted sleep classification according to the standard of the American Academy of Sleep Medicine: validation study of the AASM version of the Somnolyzer 24 × 7.

BACKGROUND: In 2007, the AASM Manual for the Scoring of Sleep and Associated Events was published by the American Academy of Sleep Medicine (AASM). Concerning the visual classification of sleep stages, these new rules are intended to replace the rules by Rechtschaffen and Kales (R&K). METHODS: We adapted the automatic R&K sleep scoring system Somnolyzer 24 × 7 to comply with the AASM rules and subsequently performed a validation study based on 72 polysomnographies from the Siesta database (56 healthy subjects, 16 patients, 38 females, 34 males, aged 21-86 years). Scorings according to the AASM rules were performed manually by experienced sleep scorers and semi-automatically by the AASM version of the Somnolyzer. Manual scorings and Somnolyzer reviews were performed independently by at least 2 out of 8 experts from 4 sleep centers. RESULTS: In the quality control process, sleep experts corrected 4.8 and 3.7% of the automatically assigned epochs, resulting in a reliability between 2 Somnolyzer-assisted scorings of 99% (Cohen's kappa: 0.99). In contrast, the reliability between the 2 manual scorings was 82% (kappa: 0.76). The agreement between the 2 Somnolyzer-assisted and the 2 visual scorings was between 81% (kappa: 0.75) and 82% (kappa: 0.76). CONCLUSION: The AASM version of the Somnolyzer revealed an agreement between semi-automated and human expert scoring comparable to that published for the R&K version with a validity comparable to that of human experts, but with a reliability close to 1, thereby reducing interrater variability as well as scoring time to a minimum.

Sleep classification according to AASM and Rechtschaffen and Kales: effects on sleep scoring parameters of children and adolescents.

Recently, the new American Academy of Sleep Medicine (AASM) rules and the old Rechtschaffen and Kales (R&K) criteria for sleep scoring have been shown to produce significantly different results in adults. The aim of this study was to describe in detail such differences in a group of normal children. Polysomnographic recordings from 45 healthy children (18 females and 27 males) aged between 3 and 16 years were scored following both systems and the results compared. Several significant differences between the two scoring systems were found: N1 (AASM) was significantly higher than S1 (R&K) while Stages N2 and R (AASM) were significantly smaller than S2 and rapid eye movement (R&K). The Kendall Tau correlation coefficient revealed a relatively low concordance between the two systems for the scoring of number of stage shifts per hour, minutes and percentage of Stage N1/S1, and of a percentage of Stage S2/N2. The significant differences between R&K and AASM scoring systems suggest taking some caution in adopting the new scoring criteria in children; these might be shown to be potentially useful if careful selection of the appropriate indicators derived from this new method is carried out, such as the percentage of N1 and the number of stage shifts, which are measures very sensitive to the occurrence of arousals in the new AASM system.

Sleep classification according to AASM and Rechtschaffen & Kales: effects on sleep scoring parameters.

STUDY OBJECTIVE: To investigate differences between visual sleep scoring according to the classification developed by Rechtschaffen and Kales (R&K, 1968) and scoring based on the new guidelines of the American Academy of Sleep Medicine (AASM, 2007). DESIGN: All-night polysomnographic recordings were scored visually according to the R&K and AASM rules by experienced sleep scorers. Descriptive data analysis was used to compare the resulting sleep parameters. PARTICIPANTS: Healthy subjects and patients (38 females and 34 males) aged between 21 and 86 years. INTERVENTIONS: N/A. MEASUREMENT AND RESULTS: While sleep latency and REM latency, total sleep time, and sleep efficiency were not affected by the classification standard, the time (in minutes and in percent of total sleep time) spent in sleep stage 1 (S1/N1), stage 2 (S2/N2) and slow wave sleep (S3+S4/N3) differed significantly between the R&K and the AASM classification. While light and deep sleep increased (S1 vs. N1 [+10.6 min, (+2.8%)]: P<0.01; S3+S4 vs. N3 [+9.1 min (+2.4%)]: P<0.01), stage 2 sleep decreased significantly according to AASM rules (S2 vs. N2 [-20.5 min, (-4.9%)]: P<0.01). Moreover, wake after sleep onset was significantly prolonged by approximately 4 minutes (P<0.01) according to the AASM standard. Interestingly, the effects on stage REM were age-dependent (intercept at 20 years: -7.5 min; slope: 1.6 min for 10-year age increase). No effects of sex and diagnosis were observed. CONCLUSION: The study shows significant and age-dependent differences between sleep parameters derived from conventional visual sleep scorings on the basis of R&K rules and those based on the new AASM rules. Thus, new normative data have to be established for the AASM standard.

The new AASM criteria for scoring hypopneas: impact on the apnea hypopnea index.

STUDY OBJECTIVES: To compare apnea-hypopnea indices (AHIs) derived using 3 standard hypopnea definitions published by the American Academy of Sleep Medicine (AASM); and to examine the impact of hypopnea definition differences on the measured prevalence of obstructive sleep apnea (OSA). DESIGN: Retrospective review of previously scored in-laboratory polysomnography (PSG). SETTING: Two tertiary-hospital clinical sleep laboratories. PATIENTS OR PARTICIPANTS: 328 consecutive patients investigated for OSA during a 3-month period. INTERVENTIONS: N/A. MEASUREMENTS AND RESULTS: AHIs were originally calculated using previous AASM hypopnea scoring criteria (AHI(Chicago)), requiring either >50% airflow reduction or a lesser airflow reduction with associated >3% oxygen desaturation or arousal. AHIs using the "recommended" (AHI(Rec)) and the "alternative" (AHI(Alt)) hypopnea definitions of the AASM Manual for Scoring of Sleep and Associated Events were then derived in separate passes of the previously scored data. In this process, hypopneas that did not satisfy the stricter hypopnea definition criteria were removed. For AHI(Rec), hypopneas were required to have > or =30% airflow reduction and > or =4% desaturation; and for AHI(Alt), hypopneas were required to have > or =50% airflow reduction and > or =3% desaturation or arousal. The median AHI(Rec) was approximately 30% of the median AHI(Chicago), whereas the median AHI(Alt), was approximately 60% of the AHI(Chicago), with large, AHI-dependent, patient-specific differences observed. Equivalent cut-points for AHI(Rec) and AHI(Alt), compared to AHI(Chicago) cut-points of 5, 15, and 30/h were established with receiver operator curves (ROC). These cut-points were also approximately 30% of AHI(Chicago) using AHI(Rec) and 60% of AHI(Chicago) using AHI(Alt). Failure to adjust cut-points for the new criteria would result in approximately 40% of patients previously classifled as positive for OSA using AHI(Chicago) being negative using AHI(Rec) and 25% being negative using AHI(Alt). CONCLUSIONS: This study demonstrates that using different published standard hypopnea definitions leads to marked differences in AHI. These results provide insight to clinicians and researchers in interpreting results obtained using different published standard hypopnea definitions, and they suggest that consideration should be given to revising the current scoring recommendations to include a single standardized hypopnea definition.

Polysomnographic values in children undergoing puberty: pediatric vs. adult respiratory rules in adolescents.

STUDY OBJECTIVES: Polysomnographic respiratory events in children should be scored using pediatric respiratory rules. However, due to a lack of data on adolescents, recently revised rules allow children aged 13-18 years to be scored by adult or pediatric criteria. To clarify which criteria to use, we describe the evolution of respiratory events with Tanner stage, and we compare events in children aged 13-18 years with the new American Academy of Sleep Medicine adult and pediatric respiratory rules. DESIGN: Cross-sectional SETTING: Academic hospital PARTICIPANTS: Healthy subjects aged 8-18 years recruited for research purposes. INTERVENTIONS: Physical examination to determine Tanner stage, overnight polysomnogram, and determination of sex hormones. RESULTS: Sixty-eight subjects (Tanner 1-5) were studied, mean age [SD] = 13 +/- 3 years, median apnea hypopnea index (AHI)= 0.1 (range: 0-1.2)/h. The median percentages of total sleep time (TST) with SpO2 < 92% were 0.1 (0-4.2)%, and with end-tidal CO2 > 50 torr was 0.1 (0-88.6)%. Thirty-two subjects were aged 13-18 years, (Tanner 3-5). The difference between AHI scored by pediatric (median = 0 [0-0.9]/h) and adult (median = 0 [0 - 0.5]/h) criteria was statistically significant (P = 0.043), but not clinically relevant. CONCLUSIONS: Respiratory events in normal children aged 8-18 years are rare and unrelated to Tanner stage. Adult or pediatric respiratory rules can be used for scoring polysomnograms in asymptomatic subjects approaching adulthood. Further studies are needed in symptomatic children within this age group.

An automatic method for the recognition and classification of the A-phases of the cyclic alternating pattern.

OBJECTIVES: The aim of this research has been to introduce an automatic method, simple from the mathematical and computational points of view, for the recognition and classification of the A-phases of the cyclic alternating pattern. METHODS: The automatic method was based on the computation of 5 descriptors, which were derived from the EEG signal and were able to provide a meaningful data reduction. Each of them corresponded to a different frequency band. RESULTS: The computation of these descriptors, followed by the introduction of two suitable thresholds and of simple criteria for logical discrimination, provided results which were in good agreement with those obtained with visual analysis. The method was versatile and could be applied to the study of other important microstructure phenomena by means of very small adaptations. CONCLUSIONS: The simplicity of the method leads to a better understanding and a more precise definition of the visual criteria for the recognition and classification of the microstructure phenomena.

Sleep studies. Current techniques and future trends.

Diagnosis of obstructive sleep apnea has been termed a laboratory diagnosis rather than a clinical diagnosis because one may not be able to make the diagnosis based on the history and physical examination alone. The polysomnogram was developed to give clinicians and researchers objective data on physiologic events occurring during the patient's sleep. From this, obstructive breathing patterns can be diagnosed and if pathologic, appropriate treatment can be instituted. Although the polysomnogram has been the gold standard for diagnosis for more than two decades, it is an expensive and time-consuming procedure. Current technologies for polysomnogram are reviewed, as well as proposals for alternatives that may be more cost and time effective.

Neonatal encephalopathies as classified by EEG-sleep criteria: severity and timing based on clinical/pathologic correlations.

Neonatal encephalopathies can be characterized in functional terms using electroencephalography. Severity of an encephalopathic state can also be estimated by electrographic interpretation independent of the time of disease process onset. Moderately or markedly abnormal electroencephalographic patterns on serial studies are highly correlated with neurologic sequelae in survivors. Electroencephalography is rarely pathognomonic or specific in determining when a condition initially occurred. However, electroencephalographic abnormalities are associated with different clinical situations, and brain lesions documented on neuroimaging or with postmortem neuropathologic examination are observed in infants with certain abnormal electrographic patterns. When interpreted in the context of history, clinical findings, and other laboratory information, the neurophysiologic studies augment the understanding of both the severity and timing of an encephalopathic state.

Towards automated sleep classification in infants using symbolic and subsymbolic approaches.

The paper addresses the problem of automatic sleep classification. A special effort is made to find a method of extracting reasonable descriptions of the individual sleep stages from sample measurements of EGG, EMG, EOG, etc., and from a classification of these measurements provided by an expert. The method should satisfy three requirements: classification accuracy, interpretability of the results, and the ability to select the relevant and discard the irrelevant variables. The solution suggested in this paper consists of a combination of the subsymbolic algorithm LVQ with the symbolic decision tree generator ID3. Results demonstrating the feasibility and utility of our approach are also presented.