A novel machine learning unsupervised algorithm for sleep/wake identification using actigraphy

ABSTRACT

Actigraphy is widely used in sleep studies but lacks a universal unsupervised algorithm for sleep/wake identification. An unsupervised algorithm is useful in large-scale population studies and in cases where polysomnography (PSG) is unavailable, as it does not require sleep outcome labels to train the model but utilizes information solely contained in actigraphy to learn sleep and wake characteristics and separate the two states. In this study, we proposed a machine learning unsupervised algorithm based on the Hidden Markov Model (HMM) for sleep/wake identification. The proposed algorithm is also an individualized approach that takes into account individual variabilities and analyzes each individual actigraphy profile separately to infer sleep and wake states. We used Actiwatch and PSG data from 43 individuals in the Multi-Ethnic Study of Atherosclerosis study to evaluate the method performance. Epoch-by-epoch comparisons and sleep variable comparisons were made between our algorithm, the unsupervised algorithm embedded in the Actiwatch software (AS), and the pre-trained supervised UCSD algorithm. Using PSG as the reference, the accuracy was 85.7% for HMM, 84.7% for AS, and 85.0% for UCSD. The sensitivity was 99.3%, 99.7%, and 98.9% for HMM, AS, and UCSD, respectively, and the specificity was 36.4%, 30.0%, and 31.7%, respectively. The Kappa statistic was 0.446 for HMM, 0.399 for AS, and 0.311 for UCSD, suggesting fair to moderate agreement between PSG and actigraphy. The Bland–Altman plots further show that the total sleep time, sleep latency, and sleep efficiency estimates by HMM were closer to PSG with narrower 95% limits of agreement than AS and UCSD. All three methods tend to overestimate sleep and underestimate wake compared to PSG. Our HMM approach is also able to differentiate relatively active and sedentary individuals by quantifying variabilities in activity counts: individuals with higher estimated activity variabilities tend to show more frequent sedentary behaviors. Our unsupervised data-driven HMM algorithm achieved better performance than the commonly used Actiwatch software algorithm and the pre-trained UCSD algorithm. HMM can help expand the application of actigraphy in cases where PSG is hard to acquire and supervised methods cannot be trained. In addition, the estimated HMM parameters can characterize individual activity patterns and sedentary tendencies that can be further utilized in downstream analysis.     

Direct comparison of two actigraphy devices with polysomnographically recorded naps in healthy young adults

The last 20 yrs have seen a marked increase in studies utilizing actigraphy in free-living environments. The aim of the present study is to directly compare two commercially available actigraph devices with concurrent polysomnography (PSG) during a daytime nap in healthy young adults. Thirty healthy young adults, ages 18–31 (mean 20.77 yrs, SD 3.14 yrs) simultaneously wore AW-64 and GT3X+ devices during a polysomnographically recorded nap. Mann-Whitney U (M-U) test, intraclass correlation coefficients, and Bland-Altman statistic were used to compare total sleep time (TST), sleep onset latency (SOL), wake after sleep onset (WASO), and sleep efficiency (SE) between the two actigraphs and PSG. Epoch-by-epoch (EBE) agreement was calculated to determine accuracy, sensitivity, specificity, predictive values for sleep (PVS) and wake (PVW), and kappa and prevalence- and bias-adjusted kappa (PABAK) coefficients. All frequency settings provided by the devices were examined. For both actigraphs, EBE analysis found accuracy, sensitivity, specificity, PVS, and PVW comparable to previous reports of other similar devices. Kappa and PABAK coefficients showed moderate to high agreement with PSG depending on device settings. The GT3X+ overestimated TST and SE, and underestimated SOL and WASO, whereas no significant difference was found between AW-64 and PSG. However, GT3X+ showed overall better EBE agreements to PSG than AW-64. We conclude that both actigraphs are valid and reliable devices for detecting sleep/wake diurnal patterns. The choice between devices should be based on several parameters as reliability, cost of the device, scoring algorithm, target population, experimental condition, and aims of the study (e.g., sleep and/or physical activity). (Author correspondence: smednick@ucr.edu)     

Feature selection for sleep/wake stages classification using data driven methods

This paper focuses on the problem of selecting relevant features extracted from human polysomnographic (PSG) signals to perform accurate sleep/wake stages classification. Extraction of various features from the electroencephalogram (EEG), the electro-oculogram (EOG) and the electromyogram (EMG) processed in the frequency and time domains was achieved using a database of 47 night sleep recordings obtained from healthy adults in laboratory settings. Multiple iterative feature selection and supervised classification methods were applied together with a systematic statistical assessment of the classification performances. Our results show that using a simple set of features such as relative EEG powers in five frequency bands yields an agreement of 71% with the whole database classification of two human experts. These performances are within the range of existing classification systems. The addition of features extracted from the EOG and EMG signals makes it possible to reach about 80% of agreement with the expert classification. The most significant improvement on classification accuracy is obtained on NREM sleep stage I, a stage of transition between sleep and wakefulness.     

Evaluation of a consumer fitness-tracking device to assess sleep in adults

Wearable fitness-tracker devices are becoming increasingly available. We evaluated the agreement between Jawbone UP and polysomnography (PSG) in assessing sleep in a sample of 28 midlife women. As shown previously, for standard actigraphy, Jawbone UP had high sensitivity in detecting sleep (0.97) and low specificity in detecting wake (0.37). However, it showed good overall agreement with PSG with a maximum of two women falling outside Bland–Altman plot agreement limits. Jawbone UP overestimated PSG total sleep time (26.6?±?35.3?min) and sleep onset latency (5.2?±?9.6?min), and underestimated wake after sleep onset (31.2?±?32.3?min) (p’s?<?0.05), with greater discrepancies in nights with more disrupted sleep. The low-cost and wide-availability of these fitness-tracker devices may make them an attractive alternative to standard actigraphy in monitoring daily sleep–wake rhythms over several days.     

Assessing Sleep Disturbance in Low Back Pain: The Validity of Portable Instruments

Although portable instruments have been used in the assessment of sleep disturbance for patients with low back pain (LBP), the accuracy of the instruments in detecting sleep/wake episodes for this population is unknown. This study investigated the criterion validity of two portable instruments (Armband and Actiwatch) for assessing sleep disturbance in patients with LBP. 50 patients with LBP performed simultaneous overnight sleep recordings in a university sleep laboratory. All 50 participants were assessed by Polysomnography (PSG) and the Armband and a subgroup of 33 participants wore an Actiwatch. Criterion validity was determined by calculating epoch-by-epoch agreement, sensitivity, specificity and prevalence and bias- adjusted kappa (PABAK) for sleep versus wake between each instrument and PSG. The relationship between PSG and the two instruments was assessed using intraclass correlation coefficients (ICC 2, 1). The study participants showed symptoms of sub-threshold insomnia (mean ISI = 13.2, 95% CI = 6.36) and poor sleep quality (mean PSQI = 9.20, 95% CI = 4.27). Observed agreement with PSG was 85% and 88% for the Armband and Actiwatch. Sensitivity was 0.90 for both instruments and specificity was 0.54 and 0.67 and PABAK of 0.69 and 0.77 for the Armband and Actiwatch respectively. The ICC (95%CI) was 0.76 (0.61 to 0.86) and 0.80 (0.46 to 0.92) for total sleep time, 0.52 (0.29 to 0.70) and 0.55 (0.14 to 0.77) for sleep efficiency, 0.64 (0.45 to 0.78) and 0.52 (0.23 to 0.73) for wake after sleep onset and 0.13 (−0.15 to 0.39) and 0.33 (−0.05 to 0.63) for sleep onset latency, for the Armband and Actiwatch, respectively. The findings showed that both instruments have varied criterion validity across the sleep parameters from excellent validity for measures of total sleep time, good validity for measures of sleep efficiency and wake after onset to poor validity for sleep onset latency.     

Newly developed waist actigraphy and its sleep/wake scoring algorithm

The purpose of this study was to formulate an algorithm for assessing sleep/waking from activity intensities measured with a waist-worn actigraphy, the Lifecorder PLUS (LC; Suzuken Co. Ltd., Nagoya, Japan), and to test the validity of the algorithm. The study consisted of 31 healthy subjects (M/F = 20/11, mean age 31.7 years) who underwent one night of simultaneous measurement of activity intensity by LC and polysomnography (PSG). A sleep(S)/wake(W) scoring algorithm based on a linear model was determined through discriminant analysis of activity intensities measured by LC over a total of 235 h and 56 min and the corresponding PSG-based S/W data. The formulated S/W scoring algorithm was then used to score S/W during the monitoring epochs (2 min each, 7078 epochs in total) for each subject. The mean agreement rate with the corresponding PSG-based S/W data was 86.9%, with a mean sensitivity (sleep detection) of 89.4% and mean specificity (wakefulness detection) of 58.2%. The agreement rates for the individual stages of sleep were 60.6% for Stage 1, 89.3% for Stage 2, 99.2% for Stage 3 + 4, and 90.1% for Stage REM. These results demonstrate that sleep/wake activity in young to middle-aged healthy subjects can be assessed with a reliability comparable to that of conventional actigraphy through LC waist actigraphy and the optimal S/W scoring algorithm.

     

Sleep-Wake Evaluation from Whole-Night Non-Contact Audio Recordings of Breathing Sounds

DesignWhole-night breathing sounds (using ambient microphone) and polysomnography (PSG) were simultaneously collected at a sleep laboratory (mean recording time 7.1 hours). A set of acoustic features quantifying breathing pattern were developed to distinguish between sleep and wake epochs (30 sec segments). Epochs (n = 59,108 design study and n = 68,560 validation study) were classified using AdaBoost classifier and validated epoch-by-epoch for sensitivity, specificity, positive and negative predictive values, accuracy, and Cohen''s kappa. Sleep quality parameters were calculated based on the sleep/wake classifications and compared with PSG for validity.SettingUniversity affiliated sleep-wake disorder center and biomedical signal processing laboratory.PatientsOne hundred and fifty patients (age 54.0±14.8 years, BMI 31.6±5.5 kg/m2, m/f 97/53) referred for PSG were prospectively and consecutively recruited. The system was trained (design study) on 80 subjects; validation study was blindly performed on the additional 70 subjects.ConclusionsThis study provides evidence that sleep-wake activity and sleep quality parameters can be reliably estimated solely using breathing sound analysis. This study highlights the potential of this innovative approach to measure sleep in research and clinical circumstances.     

A validation study of Fitbit Charge 2™ compared with polysomnography in adults

We evaluated the performance of a consumer multi-sensory wristband (Fitbit Charge 2?), against polysomnography (PSG) in measuring sleep/wake state and sleep stage composition in healthy adults.

In-lab PSG and Fitbit Charge 2? data were obtained from a single overnight recording at the SRI Human Sleep Research Laboratory in 44 adults (19—61 years; 26 women; 25 Caucasian). Participants were screened to be free from mental and medical conditions. Presence of sleep disorders was evaluated with clinical PSG. PSG findings indicated periodic limb movement of sleep (PLMS, > 15/h) in nine participants, who were analyzed separately from the main group (n = 35). PSG and Fitbit Charge 2? sleep data were compared using paired t-tests, Bland–Altman plots, and epoch-by-epoch (EBE) analysis.

In the main group, Fitbit Charge 2? showed 0.96 sensitivity (accuracy to detect sleep), 0.61 specificity (accuracy to detect wake), 0.81 accuracy in detecting N1+N2 sleep (“light sleep”), 0.49 accuracy in detecting N3 sleep (“deep sleep”), and 0.74 accuracy in detecting rapid-eye-movement (REM) sleep. Fitbit Charge 2? significantly (p < 0.05) overestimated PSG TST by 9 min, N1+N2 sleep by 34 min, and underestimated PSG SOL by 4 min and N3 sleep by 24 min. PSG and Fitbit Charge 2? outcomes did not differ for WASO and time spent in REM sleep. No more than two participants fell outside the Bland–Altman agreement limits for all sleep measures. Fitbit Charge 2? correctly identified 82% of PSG-defined non-REM–REM sleep cycles across the night. Similar outcomes were found for the PLMS group.

Fitbit Charge 2? shows promise in detecting sleep-wake states and sleep stage composition relative to gold standard PSG, particularly in the estimation of REM sleep, but with limitations in N3 detection. Fitbit Charge 2? accuracy and reliability need to be further investigated in different settings (at-home, multiple nights) and in different populations in which sleep composition is known to vary (adolescents, elderly, patients with sleep disorders).     

Changes in EMG signals for the muscle tibialis anterior while running barefoot or with shoes resolved by non-linearly scaled wavelets

The purpose of this project was to study the EMG pattern of the tibialis anterior muscle in heel-toe running. Specifically, EMG changes in time, intensity and frequency shortly before and after heel-strike were addressed using an EMG-specific non-linearly scaled wavelets analysis. This method allowed extracting the time, intensity and frequency information inherent in the EMG signal at any time. The EMG signals of 40 male subjects were recorded for running barefoot and with shoes. The results confirmed that the pre-heel-strike EMG activities were typically seen at higher EMG frequencies (60-270Hz) while the post-heel-strike EMG activities resulted in lower frequency signals (10-90Hz). The timing of the pre-heel-strike EMG activities was not influenced by the used shoe conditions. The timing of the post-heel-strike EMG activities was significantly delayed when wearing shoes. The intensity of the pre-heel-strike muscle activity increased compared to the post-heel-strike one when wearing shoes. One can conclude that the activity of the tibialis anterior adjusts specifically to exterior conditions. The frequency shift between pre- and post heel-strike muscle activity were discussed with respect to activation of different motor units.     

Muscle activation of patients suffering from asymmetric ankle osteoarthritis during isometric contractions and level walking – A time–frequency analysis

Asymmetric osteoarthritis (OA) is a common type of OA in the ankle joint. OA also influences the muscles surrounding a joint, however, little is known about the muscle activation in asymmetric ankle OA. Therefore, the aim of this study was to characterize the patients’ muscle activation during isometric ankle torque measurements and level walking. Surface electromyography (EMG) was measured of gastrocnemius medialis (GM) and lateralis (GL), soleus (SO), tibialis anterior (TA), and peroneus longus (PL) in 12 healthy subjects and 12 ankle OA patients. To obtain time and frequency components of the EMG power a wavelet transformation was performed. Furthermore, entropy was introduced to characterize the homogeneity of the wavelet patterns.Patients produced lower plantar- and dorsiflexion torques and their TA wavelet spectrum was shifted towards lower frequencies. While walking, the patients’ muscles were active with a lower intensity and over a broader time–frequency region. In contrast to controls and varus OA patients, maximal GM activity of valgus OA patients lagged behind the activity of GL and SO. In both tasks, PL of the valgus patients contained more low frequency power. The results of this study will help to assess whether surgical interventions of ankle OA can reestablish the muscle activation patterns.     

Circadian regulation gene polymorphisms are associated with sleep disruption and duration,and circadian phase and rhythm in adults with HIV

Genes involved in circadian regulation, such as circadian locomotor output cycles kaput [CLOCK], cryptochrome [CRY1] and period [PER], have been associated with sleep outcomes in prior animal and human research. However, it is unclear whether polymorphisms in these genes are associated with the sleep disturbances commonly experienced by adults living with human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS). Thus, the purpose of this study was to describe polymorphisms in selected circadian genes that are associated with sleep duration or disruption as well as the sleep–wake rhythm strength and phase timing among adults living with HIV/AIDS. A convenience sample of 289 adults with HIV/AIDS was recruited from HIV clinics and community sites in the San Francisco Bay Area. A wrist actigraph was worn for 72?h on weekdays to estimate sleep duration or total sleep time (TST), sleep disruption or percentage of wake after sleep onset (WASO) and several circadian rhythm parameters: mesor, amplitude, the ratio of mesor to amplitude (circadian quotient), and 24-h autocorrelation. Circadian phase measures included clock time for peak activity (acrophase) from actigraphy movement data, and bed time and final wake time from actigraphy and self-report. Genotyping was conducted for polymorphisms in five candidate genes involved in circadian regulation: CLOCK, CRY1, PER1, PER2 and PER3. Demographic and clinical variables were evaluated as potential covariates. Interactions between genotype and HIV variables (i.e. viral load, years since HIV diagnosis) were also evaluated. Controlling for potentially confounding variables (e.g. race, gender, CD4+ T-cell count, waist circumference, medication use, smoking and depressive symptoms), CLOCK was associated with WASO, 24-h autocorrelation and objectively-measured bed time; CRY1 was associated with circadian quotient; PER1 was associated with mesor and self-reported habitual wake time; PER2 was associated with TST, mesor, circadian quotient, 24-h autocorrelation and bed and wake times; PER3 was associated with amplitude, 24-h autocorrelation, acrophase and bed and wake times. Most of the observed associations involved a significant interaction between genotype and HIV. In this chronic illness population, polymorphisms in several circadian genes were associated with measures of sleep disruption and timing. These findings extend the evidence for an association between genetic variability in circadian regulation and sleep outcomes to include the sleep–wake patterns experienced by adults living with HIV/AIDS. These results provide direction for future intervention research related to circadian sleep–wake behavior patterns.     

Alterations of the characteristics of the circadian rest-activity rhythm of cancer in-patients

The aim of the present study was to evaluate the characteristics of the circadian rest-activity rhythm of cancer patients. Thirty-one in-patients, consisting of 19 males and 12 females, were randomly selected from the Regional Cancer Center, Pandit Jawaharlal Nehru Medical College, Raipur, India. The rest-activity rhythm was studied non-invasively by wrist actigraphy, and compared with 35 age-matched apparently healthy subjects (22 males and 13 females). All subjects wore an Actiwatch (AW64, Mini Mitter Co. Inc., USA) for at least 4-7 consecutive days. Fifteen-second epoch length was selected for gathering actigraphy data. In addition, several sleep parameters, such as time in bed, assumed sleep, actual sleep time, actual wake time, sleep efficiency, sleep latency, sleep bouts, wake bouts, and fragmentation index, were also recorded. Data were analyzed using several statistical techniques, such as cosinor rhythmometry, spectral analysis, ANOVA, Duncan's multiple-range test, and t-test. Dichotomy index (I相似文献   

Longitudinal Study of Mothers' Sleep-Wake Behaviors and Circadian Time Patterns from Late Pregnancy to Postpartum - Monitoring of Wrist Actigraphy and Sleep Logs

This longitudinal study investigated sleep-wake behavior patterns during and after pregnancy, using an actimeter worn on the non-dominant wrist and a sleep log. Records were obtained from ten mothers, from the 34th week of gestation until the 15th week postpartum. Ten non-pregnant women were used as a control group, data being collected from them for 2 weeks. The sleep-wake behavior after delivery, obtained from wrist actigraphy, was greater in the postpartum period. Total sleep time, sleep efficiency, and circadian amplitude decreased in the weeks immediately following parturition, but wake after sleep onset increased. Subsequently, all the sleep and circadian variables improved slightly, but they had not returned to the levels of the non-pregnant control group even by the 15th postpartum week. The length of daytime naps increased, in order to make up for nocturnal sleep deprivation when the number of awakenings during nighttime had increased. There were significant positive correlations between total sleep time, sleep efficiency, wake after sleep onset, and the length of daytime naps, but the numbers of awakenings at night and daytime naps did not show this correlation. The total sleep time indicated by sleep logs tended to be greater than that indicated by actigraphy, but wake after sleep onset tended to be underestimated by the sleep logs. The implications of these results are discussed.     

Longitudinal Study of Mothers’ Sleep-Wake Behaviors and Circadian Time Patterns from Late Pregnancy to Postpartum – Monitoring of Wrist Actigraphy and Sleep Logs

This longitudinal study investigated sleep-wake behavior patterns during and after pregnancy, using an actimeter worn on the non-dominant wrist and a sleep log. Records were obtained from ten mothers, from the 34th week of gestation until the 15th week postpartum. Ten non-pregnant women were used as a control group, data being collected from them for 2 weeks. The sleep-wake behavior after delivery, obtained from wrist actigraphy, was greater in the postpartum period. Total sleep time, sleep efficiency, and circadian amplitude decreased in the weeks immediately following parturition, but wake after sleep onset increased. Subsequently, all the sleep and circadian variables improved slightly, but they had not returned to the levels of the non-pregnant control group even by the 15th postpartum week. The length of daytime naps increased, in order to make up for nocturnal sleep deprivation when the number of awakenings during nighttime had increased. There were significant positive correlations between total sleep time, sleep efficiency, wake after sleep onset, and the length of daytime naps, but the numbers of awakenings at night and daytime naps did not show this correlation. The total sleep time indicated by sleep logs tended to be greater than that indicated by actigraphy, but wake after sleep onset tended to be underestimated by the sleep logs. The implications of these results are discussed.     

Diaphragm long-term facilitation following acute intermittent hypoxia during wakefulness and sleep

Acute intermittent hypoxia (AIH) elicits a form of respiratory plasticity known as long-term facilitation (LTF). Here, we tested four hypotheses in unanesthetized, spontaneously breathing rats using radiotelemetry for EEG and diaphragm electromyography (Dia EMG) activity: 1) AIH induces LTF in Dia EMG activity; 2) diaphragm LTF (Dia LTF) is more robust during sleep vs. wakefulness; 3) AIH (or repetitive AIH) disrupts natural sleep-wake architecture; and 4) preconditioning with daily AIH (dAIH) for 7 days enhances Dia LTF. Sleep-wake states and Dia EMG were monitored before (60 min), during, and after (60 min) AIH (10, 5-min hypoxic episodes, 5-min normoxic intervals; n = 9), time control (continuous normoxia, n = 8), and AIH following dAIH preconditioning for 7 days (n = 7). Dia EMG activities during quiet wakefulness (QW), rapid eye movement (REM), and non-REM (NREM) sleep were analyzed and normalized to pre-AIH values in the same state. During NREM sleep, diaphragm amplitude (25.1 ± 4.6%), frequency (16.4 ± 4.7%), and minute diaphragm activity (amplitude × frequency; 45.2 ± 6.6%) increased above baseline 0-60 min post-AIH (all P < 0.05). This Dia LTF was less robust during QW and insignificant during REM sleep. dAIH preconditioning had no effect on LTF (P > 0.05). We conclude that 1) AIH induces Dia LTF during NREM sleep and wakefulness; 2) Dia LTF is greater in NREM sleep vs. QW and is abolished during REM sleep; 3) AIH and repetitive AIH disrupt natural sleep patterns; and 4) Dia LTF is unaffected by dAIH. The capacity for plasticity in spinal pump muscles during sleep and wakefulness suggests an important role in the neural control of breathing.     

Time-frequency analysis of surface electromyographic signals during fatiguing isokinetic muscle actions

The purpose of this study was to use a wavelet analysis designed specifically for electromyography (EMG) signals in combination with a trend plot to examine changes in EMG intensity patterns during maximal, fatiguing isokinetic muscle actions. Eleven men (mean ± SD age = 22.4 ± 1.1 years) and 7 women (mean ± SD age = 22.7 ± 2.1 years) performed 50 consecutive maximal concentric isokinetic muscle actions of the dominant leg extensors at a velocity of 180°·s(-1). During each muscle action, a bipolar surface EMG signal was detected from the vastus lateralis. All signals were then processed with a wavelet analysis designed specifically for EMG signals, which resulted in EMG intensity patterns. The patterns for each subject were then analyzed with a trend plot, which provided information regarding the changes that occurred because of fatigue. The results indicated that for all the 18 subjects, the EMG intensity patterns moved in a predictable manner in pattern space, but the changes to the patterns were different for each subject. These findings reflect the complex changes that occur in the EMG signal during fatigue. These changes cannot be characterized fully with a single amplitude and center frequency parameter and can be useful for athletes and coaches who need to track the fatigue status of individual muscles.     

Effects of Neuromuscular Electrical Stimulation on the Masticatory Muscles and Physiologic Sleep Variables in Adults with Cerebral Palsy: A Novel Therapeutic Approach

Cerebral palsy (CP) is a term employed to define a group of non-progressive neuromotor disorders caused by damage to the immature or developing brain, with consequent limitations regarding movement and posture. CP may impair orapharygeal muscle tone, leading to a compromised chewing function and to sleep disorders (such as obstructive sleep apnea). Thirteen adults with CP underwent bilateral masseter and temporalis neuromuscular electrical stimulation (NMES) therapy. The effects on the masticatory muscles and sleep variables were evaluated using electromyography (EMG) and polysomnography (PSG), respectively, prior and after 2 months of NMES. EMG consisted of 3 tests in different positions: rest, mouth opening and maximum clenching effort (MCE). EMG values in the rest position were 100% higher than values recorded prior to therapy for all muscles analyzed (p < 0.05); mean mouth opening increased from 38.0 ± 8.0 to 44.0 ± 10.0 cm (p = 0.03). A significant difference in MCE was found only for the right masseter. PSG revealed an improved in the AHI from 7.2±7.0/h to 2.3±1.5/h (p < 0.05); total sleep time improved from 185 min to 250 min (p = 0.04) and minimun SaO2 improved from 83.6 ± 3.0 to 86.4 ± 4.0 (p = 0.04). NMES performed over a two-month period led to improvements in the electrical activity of the masticatory muscles at rest, mouth opening, isometric contraction and sleep variables, including the elimination of obstructive sleep apnea events in patients with CP.

Trial Registration

ReBEC RBR994XFS http://www.ensaiosclinicos.gov.br     

A transition-constrained discrete hidden Markov model for automatic sleep staging

ABSTRACT: BACKGROUND: Approximately one-third of the human lifespan is spent sleeping. To diagnose sleep problems, all-night polysomnographic (PSG) recordings including electroencephalograms (EEGs), electrooculograms (EOGs) and electromyograms (EMGs), are usually acquired from the patient and scored by a well-trained expert according to Rechtschaffen & Kales (R&K) rules. Visual sleep scoring is a time-consuming and subjective process. Therefore, the development of an automatic sleep scoring method is desirable. METHOD: The EEG, EOG and EMG signals from twenty subjects were measured. In addition to selecting sleep characteristics based on the 1968 R&K rules, features utilized in other research were collected. Thirteen features were utilized including temporal and spectrum analyses of the EEG, EOG and EMG signals, and a total of 158 hours of sleep data were recorded. Ten subjects were used to train the Discrete Hidden Markov Model (DHMM), and the remaining ten were tested by the trained DHMM for recognition. Furthermore, the 2-fold cross validation was performed during this experiment. RESULTS: Overall agreement between the expert and the results presented is 85.29%. With the exception of S1, the sensitivities of each stage were more than 81%. The most accurate stage was SWS (94.9%), and the least-accurately classified stage was S1 (<34%). In the majority of cases, S1 was classified as Wake (21%), S2 (33%) or REM sleep (12%), consistent with previous studies. However, the total time of S1 in the 20 all-night sleep recordings was less than 4%. CONCLUSION: The results of the experiments demonstrate that the proposed method significantly enhances the recognition rate when compared with prior studies.     

Comparison of Two Different Actigraphs with Polysomnography in Healthy Young Subjects

The present study aimed to compare two commercially available actigraphs, with a concurrent polysomnographic (PSG) recording. Twelve healthy volunteers (six women; age range 19–28 yrs) simultaneously wore the Basic Mini‐Motionlogger® and Actiwatch® for seven overnight polysomnographic recordings. Comparisons of the following sleep measures were focused on: sleep onset latency (SOL), total sleep time, wake after sleep onset, and sleep efficiency. Both devices underestimated SOL in comparison to PSG, but they had similar performance compared to PSG for the other sleep measures. A limit of the study is that the results can be only generalized to healthy young subjects.     

Comparison of two different actigraphs with polysomnography in healthy young subjects

The present study aimed to compare two commercially available actigraphs, with a concurrent polysomnographic (PSG) recording. Twelve healthy volunteers (six women; age range 19-28 yrs) simultaneously wore the Basic Mini-Motionlogger^® and Actiwatch^® for seven overnight polysomnographic recordings. Comparisons of the following sleep measures were focused on: sleep onset latency (SOL), total sleep time, wake after sleep onset, and sleep efficiency. Both devices underestimated SOL in comparison to PSG, but they had similar performance compared to PSG for the other sleep measures. A limit of the study is that the results can be only generalized to healthy young subjects.