The master´s thesis is focused on advanced scoring of sleep data, which was performed using deep neural network. Heart rate data and the movement information were used for scoring measured using an Apple Watch smartwatch. After appropriate pre-processing, this data serves as input parameters to the designed networks. The goal of the LSTM network was to classify data into either two groups for sleep and wake or into three groups for wake, Non-REM and REM. The best results were achieved by network doing classification of sleep vs. wake using the accelerometer. The statistical evaluation of this best-designed network reached the values of sensitivity 71,06 %, specificity 57,05 %, accuracy 70,01 % and F1 score 81,42 %.

Cílem této práce byl vývoj programu pro automatickou detekci arousalu v signálu spánkového EEG s použitím metod časově-frekvenční analýzy. Předmětem studie bylo 13 celonočních polysomnografických nahrávek (čtyři svody EEG, EMG, EKG a EOG), tj. celkově více než 100 hodin záznamu. Jednalo se o část dat z dřívějších výzkumných prací expertní lékařky v problematice spánku Dr. Emilie Sforzy, Ženeva, Švýcarsko, která rovněž poskytla základní hodnocení těchto dat. V záznamech bylo celkem označeno 1551 arousal událostí. Pro usnadnění výběru konkrétní metody časově-frekvenční analýzy byla následně vytvořena sada nástrojů pro vizualizaci jednotlivých signálů a jejich různých časově-frekvenčních vyjádření. S ohledem na závěry vizuální analýzy, charakter signálu EEG a efektivitu výpočetních metod byla pro analýzu vybrána waveletová transformace s mateřskou vlnkou Daubechies řádu 6. Jednotlivé svody EEG byly dekomponovány do šesti frekvenčních pásem. Z takto odvozených signálů a signálu EMG byly následně stanoveny ukazatele možné přítomnosti události arousalu. Tyto ukazatele byly dále váhovány lineárním klasifikátorem, jehož hodnoty vah byly optimalizovány pomocí genetického algoritmu. Na základě hodnoty lineárního klasifikátoru bylo rozhodnuto o přítomnosti události arousalu v daném svodě EEG – arousal byl detekován, jestliže hodnota klasifikátoru překročila danou mez na dobu více než 3 a méně než 30 vteřin. V celém záznamu pak byl arousal označen, byl-li detekován alespoň v jednom ze svodů EEG. Následně byly odvozeny míry senzitivity a selektivity detekce, jež byly rovněž základem pro stanovení fitness funkce genetického algoritmu. Pro učení genetického algoritmu byly vybrány první čtyři záznamy. Na základě takto optimalizovaných vah vznikl program pro automatickou detekci, který na celém souboru 13 záznamů dosáhl ve srovnání s expertním hodnocením míry senzitivity 76,09%, selektivity 53,26% a specificity 97,66%.

This paper addresses the problem of detecting K-complexes in sleep EEG. The study of sleep has become very essential to diagnose the brain disorders and analysis of brain activities. Since Kcomplex can have a wide variety of shapes it is very difficult to detect the K-complexes manually. In this paper, I present an automatic method for K-complexes detection based wavelet transform,TKEO and method for classification using feedforward multilayer neural network designed in Matlab. Detection performance reached the value approx. from 52,9 to 83,6 %.

The thesis is focused on automatic classification of polysomnographic signals based on various parameters in time and frequency domain. The parameters are acquired from 30 seconds long segments of EEG, EMG and EOG signals recorded during different sleep stages. The parameters used for automatic classification of sleep stages are selected according to statistical analysis. Classification is performed using the SVM method and evaluation of the success of the classification is done using sensitivity, specificity and percentage success. Classification method was implemented using Matlab.

Une mauvaise qualité de sommeil est souvent observée chez les personnes âgées. Diverses méthodes sont alors utilisées afin d’en évaluer les caractéristiques, notamment les questionnaires auto-recensés et la polysomnographie (PSG). Cependant, l’utilisation de la PSG étant dispendieuse et chronophage, le recours aux analyses par questionnaires auto-recensés en milieu clinique pourrait être privilégié et faciliterait grandement l’évaluation de la qualité du sommeil. Cette étude vise à valider la méthode par questionnaires auto-recensés lors de l’évaluation de la qualité de sommeil chez les personnes âgées et à comparer ces différents questionnaires aux mesures obtenues par une PSG. Le devis de l'étude est une une sous-analyse d’une étude d’essais cliniques randomisés. L’étude a été menée auprès de patients édentés de 65 ans et plus, recrutés entre décembre 2013 et août 2018. Les données ont été collectées à domicile avec une PSG et jumelées à divers questionnaires sur le sommeil : le questionnaire de Berlin (QB, plage de scores comprise entre 0 et 3), l’échelle de somnolence Karolinska (ESK, plage de scores 1-9), l’indice de qualité du sommeil de Pittsburgh (IQSP, plage de scores de 0 à 21), l’échelle de somnolence d’Epworth (ESE, plage de scores de 0 à 24). Les données obtenues par la PSG incluent : l'indice d'apnée-hypopnée (IAH), l'indice de désaturation en oxygène (IDO) et le pourcentage d'efficacité du sommeil. L’analyse des données a été effectuée par statistiques descriptives et tests statistiques bivariés. Chez les 130 patients ayant participé à l’acquisition de données de base (51 hommes et 79 femmes, âge moyen 75 ± 6 ans), il n'y a pas de forte corrélation entre les scores totaux de qualité du sommeil mesurés par les questionnaires et le pourcentage d'IAH, d'ODI et d'efficacité du sommeil obtenus par la PSG. Seuls deux des questionnaires utilisés ont eu une sensibilité distinctive, comportant toutefois des valeurs de seuil différentes de la norme utilisée chez les adultes : le QB et l’ESK. Nos résultats démontrent que les questionnaires auto-recensés disponibles s’avèrent limités pour le dépistage clinique chez les personnes âgées. Afin de développer un nouvel instrument de mesure fiable permettant le diagnostic d’AOS et l’évaluation de la qualité de sommeil chez cette population, il est nécessaire d’effectuer des études sur un plus grand échantillon et de créer un questionnaire spécifique destiné à la population âgée. / Poor sleep quality is often seen in the elderly. Various methods are then used to evaluate their characteristics, including self-reported questionnaires and polysomnography (PSG). However, since the use of PSG is expensive and time-consuming, the use of self-reported questionnaires in a clinical setting could be preferred and would greatly facilitate the assessment of sleep quality. This study aims to validate the self-reported questionnaire method used in the evaluation of sleep quality in the elderly and compare these different questionnaires to the measures obtained by a PSG. The study design is a sub-analysis of a randomized clinical trial study. The study was conducted in edentulous patients aged 65 years and older, recruited between December 2013 and August 2018. The data were collected at home with a PSG and combined with various sleep questionnaires: the Berlin questionnaire (BQ, range of scores from 0 to 3), the Karolinska sleepiness scale (KSS, range of scores from 1-9), the Pittsburgh Sleep Quality Index (PSQI, range of scores from 0 to 21), the Epworth sleepiness scale (ESS, range of scores from 0 to 24). Data obtained by PSG include: apnea-hypopnea index (AHI), oxygen desaturation index (ODI), and sleep efficiency percentage. Data analysis was performed by descriptive statistics and bivariate statistical tests. Of the 130 patients who participated in baseline data acquisition (51 males and 79 females, mean age 75 ± 6 years), there was no strong correlation between the total sleep quality scores measured by the questionnaires and the IAH, ODI and sleep efficiency percentage achieved by PSG. Only two of the questionnaires used had a distinctive sensitivity, but with threshold values different from the norm used in adults: QB and ESK. Our results demonstrate that self-reported questionnaires available are limited for clinical screening for the elderly. In order to develop a new reliable measuring instrument for the diagnosis of obstructive sleep apnea (OSA) and the evaluation of sleep quality in this population, it is necessary to carry out studies on a larger sample and to create a specific questionnaire intended for the elderly population.

Sleep and dreaming are important research topics. Unfortunately, the methods for researching them have several shortcomings. In-laboratory polysomnographic sleep and dream research is a costly, time-consuming and effortful endeavor, often resulting in small subject counts. Moreover, the unfamiliar sleeping environment can lead to distorted measurements as compared to the natural sleep environment at the subject’s home. Conducting sleep and dream experiments in the field by a crowd of subjects could be a solution. However, complex experiment paradigms cannot be investigated this way, because there are no tools available, which enable naive subjects to carry out complex polysomnographic studies on their own. The Traumschreiber system, which is developed and evaluated in this dissertation, offers a solution to this problem. It consists of a high-tech sleep mask and a minicomputer, and enables naive crowd subjects to perform complex polysomnographic sleep and dream experiments at home. On the one hand, it instructs the crowd subject, what to do when. On the other hand, it controls the experiment during the time the subject is asleep, analyzing the data in real-time using state-of-the art machine learning techniques. The rationale behind is to enable a big data approach to sleep and dream research, using the data recorded by a crowd of subjects for large-scale investigations about sleep and dreaming, with low costs for the researcher. After describing the development process of the Traumschreiber system, its usefulness regarding crowd-based automated polysomnographic field studies is evaluated. First, it is validated against a commercial medical poly­somnographic sleep laboratory system, demonstrating its good polysomnographic data recording capabilities – including measurements of EEG, EOG, EMG and ECG –, which enable the researcher to identify typical sleep patterns like slow waves or rapid eye movements as well as sleep stages in the recorded data. Furthermore, two field studies show, that the Traumschreiber system can be used successfully by naive subjects to conduct complex sleep experiments at their homes. This includes acoustic stimulation of the sleeping subject as well as sleep stage dependent activities of the system. The sleep staging algorithm implements a Keras/Tensorflow based neural network approach, which demonstrates the system’s readiness for state-of-the-art machine learning techniques. However, the currently used neural network is kept very simple and can determine the sleep stage not very reliably; it should be further developed and trained on more data of more subjects. The Traumschreiber system will be made available under an open source license, enabling any researcher to use, modify or further develop it. A description, how to produce arbitrarily many entities of the Traumschreiber system, is given in this dissertation and shows that one system can be produced at low costs in a short amount of time. Taken together, the Traumschreiber system is a new tool for sleep and dream research, which enables a crowd-based and machine learning-driven approach to gathering polysomnographic data from complex sleep and dream experiments.

Introduction: Les évidences scientifiques actuelles sur l’apnée du sommeil pédiatrique considérant le développement crâniofacial sont limitées, et l’identification des enfants vulnérables est difficile, car un manque d’évidence existe quant aux facteurs de risque et aux symptômes de cette population. Objectifs: Évaluer l’histoire naturelle de l’apnée du sommeil, de l’enfance à la fin de l’adolescence, afin de récolter de l’information supplémentaire quant à la persistance, la rémission et l’incidence de la maladie, tout en considérant l’impact de la croissance et du développement crâniofacial. Les objectifs secondaires de cette étude sont également d’évaluer la trajectoire des comorbidités associées à l’apnée du sommeil (comportementales et neurocognitives). Matériel et Méthode : La cohorte d’enfants initialement recrutés au CHUSJ, pour qui des troubles de sommeil étaient suspectés, ont été recontactés pour un suivi 5 ans suivant leur date initiale d’évaluation. Dix-neuf enfants ont été réévalués en complétant des examens dentaires et orthodontiques, des questionnaires (Epworth, PSQI, CPRS-R) ainsi qu’une polysomnographie à domicile. Résultats: L’âge moyen des participants adolescents à V2 étaient 12.79 ± 2.74 à. À V2, les garçons (IAH=3.28 ± 2.43) étaient atteints plus sévèrement que les filles (IAH=2.81 ± 2.02), mais de façon non significative (p=0.589). Au suivi 5 ans plus tard (V2), aucun enfant était atteint sévèrement alors que 42% étaient légèrement atteints et 22% souffraient d’une AOS modérée comparativement à V1, où 36% des enfants étaient atteints légèrement, 18% atteints de façon modérée et 18% souffraient d’apnée sévère. La prévalence à V2 est de 63.2% alors que l’incidence est de 54.5%. Un taux de de rémission de 25% a été noté. Aucune caractéristique dentaire ni squelettique a été associée à la présence et à la persistance de l’apnée du sommeil. Conclusion: Les troubles respiratoires ne sont pas nécessairement résolus à l’adolescence, et de nouveaux facteurs de risque, similaires à ceux retrouvés chez l’adulte, font émergence. La connaissance des caractéristiques crâniofaciales associées à l’AOS est essentielle pour optimiser les traitements et maximiser les récurrences. / Introduction: Current scientific studies of the history of pediatric obstructive sleep apnea that consider craniofacial development impacts remain somewhat limited, while there is incomplete global evidence for establishing risk factor and symptom algorithm that adequately identify vulnerable children. Objective: To evaluate the natural history of sleep apnea, from childhood to the end of adolescence, and gather additional information regarding the persistence, remission, and incidence of the disease while considering the impact of growth and craniofacial development. As secondary goals, the study will assess the trajectory of associated sleep apnea comorbidities (behavioral and neurocognitive). Methods: The cohort of children who were initially recruited at the CHUSJ, for whom obstructive sleep apnea was suspected, were recalled for a 5-year follow-up, according to the date of their first evaluation. A total of 19 children were seen to complete dental and orthodontic evaluations, three questionnaires (Epworth, PSQI, CPRS-R) and a home-based polysomnography. Results: The mean age of the adolescent participants at follow-up was 12.79 ± 2.74. At V2, boys (AHI=3.28 ± 2.43) were more severely affected than girls (2.81 ± 2.02) but the difference was not significant (p=0.589). At the 5-year follow-up (V2), none of the children were severely ill but 42% had mild sleep apnea while 22% had moderate apnea comparatively to V1 where 36% had mild apnea, while 18% suffered from moderate apnea and 18% were severely ill. Prevalence at V2 was 63.2% and incidence was 54.5%. Remission rate was 25%. No skeletal or dental features were associated with the presence and persistence of obstructive sleep apnea. Conclusion: Obstructive sleep disordered breathing are not necessarily resolved at adolescence. On the other hand, with the onset of adolescence, new risk factors similar to those we find in adults emerge. In order to maximize treatments outcome and minimize recurrences, knowledge of craniofacial features is essential.

REM sleep behavior disorder: Characteristics of polysomnographic and behavioral manifestations Abstract REM sleep behavior disorder (RBD) is a disease characterized by abnormal motor activity corresponding to the dream content. REM sleep without atonia (RWA) and behavioral manifestations are the main features registered by video-polysomnography (PSG). Because idiopathic RBD (iRBD) is considered as prodromal stage of synucleinopathies, the direction of current research is the search for markers of early conversion. The goal of this study was to observe the group of patients with iRBD with regard to the development of manifest neurodegenerative disease, to find and test a new polysomnographic marker of phenoconversion, to perform analysis of the movements registered by video and to quantify excessive fragmentary myoclonus (EFM), which is a frequent finding in neurodegenerative processes. A total of 55 patients with iRBD were observed for 2.3±0.7 years. The annual conversion rate was 5.5%. Mixed RWA, representing simultaneous occurrence of phasic and tonic RWA, was suggested as a new marker of phenoconversion. Converted patients showed a higher mixed RWA (p=0.009) and the ROC analysis confirmed that mixed RWA is the best predictive marker of conversion among other RWA types (AUC 0.778). An average of...

Sleep arousal is a phenomenon that affects the sleep of a large amount of people. The process of predicting and classifying arousal events is done manually with the aid of certified technologists, although some research has been done on automation using Artificial Neural Networks (ANN). This study explored how a Support Vector Machine performed(SVM) compared to an ANN on this task. Polysomnography (PSG) is a sort of sleep study which produces the data that is used in classifying sleep disorders. The PSG-data used in this thesis consists of 13 wave forms sampled at or resampled at 200Hz. There were samples from 994 patients totalling approximately 6.98 1010 data points, processing this amount of data is time consuming and presents a challenge. 2000 points of each signal was used in the construction of the data set used for the models. Extracted features included: Median, Max, Min, Skewness, Kurtosis, Power of EEG-band frequencies and more. Recursive feature elimination was used in order to select the best amount of extracted features. The extracted data set was used to train two ”out of the box” classifiers and due to memory issues the testing had to be split in four batches. When taking the mean of the four tests, the SVM scored ROC AUC of 0,575 and the ANN 0.569 respectively. As the difference in the two results was very modest it was not possible to conclude that either model was better suited for the task at hand. It could however be concluded that SVM can perform as well as ANN on PSG-data. More work has to bee done on feature extraction, feature selection and the tuning of the models for PSG-data to conclude anything else. Future thesis work could include research questions as ”Which features performs best for a SVM in the prediction of Sleep arousals on PSG-data” or ”What feature selection technique performs best for a SVM in the prediction of Sleep arousals on PSG-data”, etc. / Sömnstörningar är en samling hälsotillstånd som påverkar sömnkvaliteten hos en stor mängd människor. Ett exempel på en sömnstörning är sömnapne. Detektion av dessa händelser är idag en manuell uppgift utförd av certifierade teknologer, det har dock på senare tid gjorts studier som visar att Artificella Neurala Nätverk (ANN) klarar att detektera händelserna med stor träffsäkerhet. Denna studie undersöker hur väl en Support Vector Machine (SVM) kan detektera dessa händelser jämfört med en ANN. Datat som används för att klassificera sömnstörningar kommer från en typ av sömnstudie kallad polysomnografi (PSG). Den PSG-data som används i denna avhandling består av 13 vågformer där 12 spelats in i 200Hz och en rekonstruerats till 200Hz. Datan som används i denna avhandling innehåller inspelningar från 994 patienter, vilket ger totalt ungefär·6.98 1010 datapunkter. Att behandla en så stor mängd data var en utmaning. 2000 punkter från vare vågform användes vid konstruktionen av det dataset som användes för modellerna. De attribut som extraherades innehöll bland annat: Median, Max, Min, Skewness, Kurtosis, amplitud av EEG-bandfrekvenser m.m. Metoden Recursive Feature Elimination användes för att välja den optimala antalet av de bästa attributen. Det extraherade datasetet användes sedan för att träna två standard-konfigurerade modeller, en SVM och en ANN. På grund av en begräning av arbetsminne så var vi tvungna att dela upp träningen och testandet i fyra segment. Medelvärdet av de fyra testen blev en ROC AUC på 0,575 för en SVM, respektive 0,569 för ANN. Eftersom skillnaden i de två resultaten var väldigt marginella kunde vi inte dra slutsatsen att endera modellen var bättre lämpad för uppgiften till hands. Vi kan dock dra slutsatsen att en SVM kan prestera lika väl som ANN på PSG-data utan konfiguration. Mer arbete krävs inom extraheringen av attributen, attribut-eliminationen och justering av modellerna. Framtida avhandlingar skulle kunna göras med frågeställningarna: “Vilka attributer fungerar bäst för en SVM inom detektionen av sömnstörningar på PSG-data” eller ”Vilken teknik för attribut-elimination fungerar bäst för en SVM inom detektionen av sömnstörningar på PSG-data”, med mera.

Contexte : Le traitement des informations externes étant toujours présent pendant le sommeil, présenter des stimulations sensorielles durant le sommeil peut en améliorer la qualité. Le mouvement latéral et doux d’un lit berçant, a montré des résultats prometteurs chez l’humain et la souris, sur une seule nuit chez des bons dormeurs. Objectif : L’objectif de ce projet pilote est d’évaluer les effets préliminaires du bercement durant le sommeil (mouvement latéral, 0.25Hz) sur la qualité du sommeil pendant plusieurs nuits consécutives. Nous cherchons à tester notre lit berceur et à reproduire les effets bénéfiques trouvés dans la littérature sur la qualité du sommeil, et à déterminer si ses effets se maintiennent lors de plusieurs nuits. Méthodologies : Le sommeil de 8 jeunes bons dormeurs (24.25  3.20 ans), a été évalué objectivement (polysomnographie, PSG), et subjectivement (questionnaires) durant 6 nuits, dont 5 nuits expérimentales sur le lit berceur (3 nuits bercées (B), 2 nuits stationnaires (S)). Résultats : Lors de B1, les participants ont passé moins de temps en sommeil profond (N3; p=0.013), et plus de temps éveillé (p=0.03) que lors de S1. En B2 ils ont passé plus de temps éveillé qu’en S2 (p=0.039), moins en B2 qu’en B1 (p=0.023), sans changements entre S1 et S2 ni entre B2 et B3 (toutes p>0.05), bien que 6/8 participants ont augmenté leur %N3 et diminué leur %N1+N2 de B2 à B3. Aucune influence du bercement sur la qualité du sommeil subjective, l’humeur ou la perception du sommeil n’est ressortie (toutes p>0.05). Conclusion : Au cours des 3 nuits bercées, une habituation du dormeur au bercement semble se produire. L’absence de résultats bénéfiques viendrait des propriétés de l’accélération linéaire du moteur. / Context: Since the process of external information continues during sleep, presenting sensory stimuli during sleep can enhance its quality. The gentle lateral movement of a rocking bed has shown promising results in humans and mice during a single night with good sleepers. Objective: The objective of this pilot study is to evaluate the preliminary effects of rocking during sleep (lateral movement, 0.25Hz) on sleep quality over multiple consecutive nights. We aim to test our rocking bed and replicate the beneficial effects found in the literature on sleep quality, as well as determine if these effects persist over several nights. Methodology: The sleep of 8 young good sleepers (24.25 ± 3.20 years) was objectively assessed (polysomnography, PSG) and subjectively evaluated (questionnaires) over 6 nights, including 5 experimental nights on the rocking bed (3 rocking nights (B), 2 stationary nights (S)). Results: During B1, participants spent less time in deep sleep (N3; p=0.013) and more time awake (p=0.03) compared to S1. They spent more time awake in B2 than in S2 (p=0.039), and in B2 than in B1 (p=0.023), with no changes between S1 and S2 nor between B2 and B3 (all p>0.05). However, 6/8 participants increased their %N3 and decreased their %N1+N2 from B2 to B3. No influence of rocking on subjective sleep quality, mood, or sleep perception was observed (all p>0.05). Conclusion: Over the course of the 3 rocking nights, a habituation of the participant to the rocking movement seems to occur. The absence of beneficial results may be attributed to the linear acceleration properties of the motor.