Dissertations / Theses on the topic 'Human sleep EEG analysis'

This thesis deals with analysis and processing of the Sleep Electroencephalogram (EEG) signals. The scope of this thesis can be split into several areas. The first area is application of the Independent Component Analysis (ICA) method for EEG signal analysis. A model of EEG signal formation is proposed and conditions under which this model is valid are examined. It is shown that ICA can be used to remove non-deterministic artifacts contained in the EEG signals. The second area of interest is analysis of stationarity of the Sleep EEG signal. Methods to identify stationary signal segments and to analyze statistical properties of these stationary segments are presented. The third area of interest focuses on spectral analysis of the Sleep EEG signals. Analyses are performed that shows the processes that form particular parts of EEG signals spectrum. Also, random signals that are an integral part of the EEG signals analysis are performed. The last area of interest focuses on elimination of the transition processes that are caused by the filtering of the short EEG signal segments.

Thesis (M.S.)--Worcester Polytechnic Institute.
Keywords: association rule mining; logistic regression; statistical significance of rules; window-based association rule mining; data mining; sleep data. Includes bibliographical references (leaves 166-167).

This thesis focuses on the design, development, and exploratory analysis of a human sleep data repository. We have successfully collected comprehensive data for 1,046 sleep disorder patients and created a Terabyte-scale database system to handle it. The data for each patient was collected from the patient's medical records, and from the patient's allnight sleep study (for a total of about 0.6 Gigabytes per patient). Data collected from the patient's medical record contain more than 70 attributes, including demographic data, smoking, drinking, and exercise habits, depression and daytime sleepiness questionnaires, and overall medical history. Data collected from the patient's all-night sleep study consist of 50-55 time-series signals recorded during a period of 6-8 hours at the hospital's sleep clinic. These signals include among others an electroencephalogram, electromyogram, electrooculogram, electrocardiogram, and signals tracking blood oxygen level, body position, limb movements, snoring and blood pressure. 350 additional attributes summarize sleep related events taking place during the night long study, including sleep stages, arousals, and respiratory disturbances. Particular attention during the development of our database system was paid to a database design that effectively handles the data size and complexity, that describes the structure of sleep data in clinically meaningful terms, and that will facilitates the discovery of patterns in sleep data using machine learning algorithms. We have interfaced our database with Weka, a well known data mining system. To the best of our knowledge, our database is one of the world's largest and most comprehensive in the domain of human sleep disorders.

For sleep classification, automatic electroencephalogram (EEG) interpretation techniques are of interest because they are labour saving, in contrast to manual (visual) methods. More importantly, some automatic methods, which offer a less subjective approach, can provide additional information which it is not possible to obtain by manual analysis. An extensive literature review has been undertaken to investigate the background of automatic EEG analysis techniques. Frequency domain and time domain methods are considered and their limitations are summarised. The weakness in the R & K rules for visual classification and from which most of the automatic systems borrow heavily are discussed. A new technique - model based dynamic analysis - was developed in an attempt to classify the sleep EEG automatically. The technique comprises of two phases, these are the modelling of EEG signals and the analysis of the model's coefficients using dynamic systems theory. Three techniques of modelling EEG signals are compared: the implementation of the non-linear prediction technique of Schaffer and Tidd (1990) based on chaos theory; Kalman filters and a recursive version of a radial basis function for modelling and forecasting the EEG signals during sleep. The Kalman filter approach produced good results and this approach was used in an attempt to classify the EEG automatically. For classifying the model's (Kalman filter's) coefficients, a new technique was developed by a state-space approach. A 'state variable' was defined based on the state changes of the EEG and was shown to be correlated with the depth of sleep. Furthermore it is shown that this technique may be useful for automatic sleep staging. Possible applications include automatic staging of sleep, detection of micro-arousals, anaesthesia monitoring and monitoring the alertness of workers in sensitive or potentially dangerous environments.

The use of mobile phones is continually increasing throughout the world, with recent figures showing that there are currently more than 2 billion mobile phone users worldwide. However, despite the recognised benefits of the introduction and widespread use of mobile phone technologies, concerns regarding the potential health effects of exposure to the radiofrequency electromagnetic fields emitted by mobile phone handsets have similarly increased, leading to an increase in demand for scientific research to investigate the possibility of health effects related to the use of mobile phones. An increasing amount of radiofrequency bioeffects research related to mobile phone use has focussed on the possible effects of mobile phone exposure on human brain activity and function, particularly as the absorption of energy in the head and brain region is much higher than in other body regions, which is a direct result from the close proximity of the mobile phone to the head when in normal use. In particular, the use of sleep research has become a more widely used technique for assessing the possible effects of mobile phones on human health and wellbeing, and is particularly useful for providing important information in the establishment of possible radiofrequency bioeffects, especially in the investigation of potential changes in sleep architecture resulting from mobile phone use. A review of the previous literature showed that a number of studies have reported an increase in the electroencephalogram spectral power within the 8 � 14 Hz frequency range in both awake and sleep states following radiofrequency electromagnetic field exposure. In regards to sleep, the enhancements reported have not been entirely consistent, with some early studies failing to find an effect, while more recent studies have reported that the effect differs in terms of particular frequency range. However, in general the previous literature suggests that there is an effect of mobile phone emissions on the sleep electroencephalogram, particularly in the frequency range of sleep spindle activity. In addition to changes in spectral power, changes in other conventional sleep parameters and the production and secretion of melatonin have also been investigated, however, there has been little or no consistency in the findings of previous studies, with the majority of recent studies concluding that there is no influence of mobile phone radiofrequency fields on these parameters of sleep or melatonin. Following a detailed review of the previous research, the current study was developed with the aim to improve on previous methodological and statistical limitations, whilst also being the largest study to investigate mobile phone radiofrequency bioeffects on human sleep. The principle aims were thus to test for the immediate effects of mobile phone radiofrequency electromagnetic fields on human sleep architecture and the secretion of the pineal hormone, melatonin. The experiment included 50 participants who were randomly exposed to active and sham mobile phone exposure conditions (one week apart) for 30 minutes prior to a full night-time sleep episode. The experimental nights employed a randomised exposure schedule using a double-blind crossover design. Standard polysomnography was used to measure subsequent sleep, and in addition, participants were required to provide urine samples immediately following exposure and upon waking in the morning. A full dosimetric assessment of the exposure system was also performed in order to provide sufficient details of the exposure set-up used in the current thesis and to account for the lack of detailed dosimetric data provided in the majority of previous studies. The results of the current study suggest that acute exposure to a mobile phone prior to sleep significantly enhances electroencephalogram spectral power in the sleep spindle frequency range compared to the sham exposure condition. The current results also suggest that this mobile phone-induced enhancement in spectral power is largely transitory and does not linger throughout the night. Furthermore, a reduction in rapid eye movement sleep latency following mobile phone exposure was also found compared to the sham exposure, although interestingly, neither this change in rapid eye movement sleep latency or the enhancement in spectral power following mobile phone exposure, led to changes in the overall quality of sleep. Finally, the results regarding melatonin suggested that, overall, overnight melatonin secretion is unaffected by acute exposure to a mobile phone prior to sleep. In conclusion, the current study has confirmed that a short exposure to the radiofrequency electromagnetic fields emitted by a mobile phone handset immediately prior to sleep is sufficient to induce changes in brain activity in the initial part of sleep. The consequences or functional significance of this effect are currently unknown and it would be premature to draw conclusions about possible health consequences based on the findings of the current study.

Electroencephalographic data recorded on the human scalp can be modeled as a linear mixture of underlying dipolar source generators. The characterization of such generators is the aim of several families of signal processing methods. In this HDR we consider in several details three of such families, namely 1) EEG distributed inverse solutions, 2) diagonalization methods, including spatial filtering and blind source separation and 3) Riemannian geometry. We highlight our contributions in each of this family, we describe algorithms reporting all necessary information to make purposeful use of these methods and we give numerous examples with real data pertaining to our published studies. Traditionally only the single-subject scenario is considered; here we consider in addition the extension of some methods to the simultaneous multi-subject recording scenario. This HDR can be seen as an handbook for EEG source analysis. It will be particularly useful to students and other colleagues approaching the field.

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%.

Many research studies have analyzed the effect that whole-body vibration (WBV) has on sleep, and some have sought to use vibration to treat sleep disorders such as insomnia. It has been shown that low frequencies (f < 2Hz) are generally sleep inducing, but oscillations of this frequency are typically difficult to achieve using electromagnetic vibration drives. In the research that has been performed, optimal vibration parameters have not been determined, and the effects of multiple vibration sources vibrating at different frequencies to induce a low frequency traveling wave have not been explored. Insomnia affects millions of people worldwide, and non-pharmacological treatment options are limited. A bed excited with multiple vibration sources was used to explore beat frequency vibration as a non-pharmacological treatment for insomnia. A repeated measures design pilot study of 14 participants with mild-moderate insomnia symptom severity was conducted to determine the effects of beat frequency vibration, and traditional standing wave vibration on sleep latency and quality. Participants were monitored using high-density electroencephalography (HD-EEG). Sleep latency was compared between treatment conditions. Trends of a decrease in sleep latency due to beat frequency vibration were found (p ≤ 0.181 for AASM latency, and p ≤ 0.068 for unequivocal sleep latency). Neural complexity during wake, N1, and N2 stages were compared using Multi-Scale Sample Entropy (MSE), which demonstrated significantly lower MSE between wake and N2 stages (p ≤ 0.002). Lower MSE was found in the transition from wake to N1 stage sleep but did not reach significance (p ≤ 0.300). During N2 sleep, beat frequency vibration shows lower MSE than the control session in the left frontoparietal region. This indicates that beat frequency vibration may lead to a decrease of conscious awareness during deeper stages of sleep. Standing wave vibration caused reduced Alpha activity and increased Delta activity during wake. Beat frequency vibration caused increased Delta activity during N2 sleep. These preliminary results suggest that beat frequency vibration may help individuals with insomnia symptoms by decreasing sleep latency, by reducing their conscious awareness, and by increasing sleep drive expression during deeper stages of sleep. Standing wave vibration may be beneficial for decreasing expression of arousal and increasing expression of sleep drive during wake, implying that a dynamic vibration treatment may be beneficial. The application of vibration treatment as part of a heuristic sleep model is discussed.

Epilepsy is a neurological disorder characterized by recurrent seizures. Sleep problems can cooccur with epilepsy, and adversely affect seizure diagnosis and treatment. In fact, the relationship between sleep and seizures in individuals with epilepsy is a complex one. Seizures disturb sleep and sleep deprivation aggravates seizures. Antiepileptic drugs may also impair sleep quality at the cost of controlling seizures. In general, particular vigilance states may inhibit or facilitate seizure generation, and changes in vigilance state can affect the predictability of seizures. A clear understanding of sleep-seizure interactions will therefore benefit epilepsy care providers and improve quality of life in patients. Notable progress in neuroscience research—and particularly sleep and epilepsy—has been achieved through experimentation on animals. Experimental models of epilepsy provide us with the opportunity to explore or even manipulate the sleep-seizure relationship in order to decipher different aspects of their interactions. Important in this process is the development of techniques for modeling and tracking sleep dynamics using electrophysiological measurements. In this dissertation experimental and computational approaches are proposed for modeling vigilance dynamics and their utility demonstrated in nonepileptic control mice. The general framework of hidden Markov models is used to automatically model and track sleep state and dynamics from electrophysiological as well as novel motion measurements. In addition, a closed-loop sensory stimulation technique is proposed that, in conjunction with this model, provides the means to concurrently track and modulate 3 vigilance dynamics in animals. The feasibility of the proposed techniques for modeling and altering sleep are demonstrated for experimental applications related to epilepsy. Finally, preliminary data from a mouse model of temporal lobe epilepsy are employed to suggest applications of these techniques and directions for future research. The methodologies developed here have clear implications the design of intelligent neuromodulation strategies for clinical epilepsy therapy.

The main objective of this work was the development of a computer-based Expert Sleep Analysis Methodology (ESAM) to aid sleep care physicians in the diagnosis of pre-Parkinson's disease symptoms using polysomnogram data. ESAM is significant because it streamlines the analysis of the human sleep cycles and aids the physician in the identification, treatment, and prediction of sleep disorders. In this work four aspects of computer-based human sleep analysis were investigated: polysomnogram interpretation, pre-processing, sleep event classification, and abnormal sleep detection. A review of previous developments in these four areas is provided along with their relationship to the establishment of ESAM. Polysomnogram interpretation focuses on the ambiguities found in human polysomnogram analysis when using the rule based 1968 sleep staging manual edited by Rechtschaffen and Kales (R&K). ESAM is presented as an alternative to the R&K approach in human polysomnogram interpretation. The second area, pre-processing, addresses artifact processing techniques for human polysomnograms. Sleep event classification, the third area, discusses feature selection, classification, and human sleep modeling approaches. Lastly, abnormal sleep detection focuses on polysomnogram characteristics common to patients suffering from Parkinson's disease. The technical approach in this work utilized polysomnograms of control subjects and pre-Parkinsonian disease patients obtained from the Emory Clinic Sleep Disorders Center (ECSDC) as inputs into ESAM. The engineering tools employed during the development of ESAM included the Generalized Singular Value Decomposition (GSVD) algorithm, sequential forward and backward feature selection algorithms, Particle Swarm Optimization algorithm, k-Nearest Neighbor classification, and Gaussian Observation Hidden Markov Modeling (GOHMM). In this study polysomnogram data was preprocessed for artifact removal and compensation using band-pass filtering and the GSVD algorithm. Optimal features for characterization of polysomnogram data of control subjects and pre-Parkinsonian disease patients were obtained using the sequential forward and backward feature selection algorithms, Particle Swarm Optimization, and k-Nearest Neighbor classification. ESAM output included GOHMMs constructed for both control subjects and pre-Parkinsonian disease patients. Furthermore, performance evaluation techniques were implemented to make conclusions regarding the constructed GOHMM's reflection of the underlying nature of the human sleep cycle.

Fibromyalgia syndrome (FMS) is a disorder characterised by chronic widespread pain, non-restorative sleep, fatigue and daytime dysfunction. Occurring in 2-5% of the population, the aetiology is largely unknown. Sleep dysfunction occurs in over 90% of FMS patients. While research has shown that both the macrostructure and microstructure of sleep may be altered, there remain inconsistencies in the polysomnographic (PSG) findings, and wide variations in methodological approaches. Few studies have controlled for symptom duration or the time elapsed between diagnosis and PSG sleep assessments. In addition, while psychometric analyses have suggested a distinctive FMS psychological profile (which includes higher levels of depressive symptoms, anxiety and fatigue) few studies have simultaneously, and thoroughly examined sleep and psychological status in the same participants. A frequently reported alteration found in the sleep microstructure of FMS patients is the alpha-delta sleep anomaly, characterised by an increase in alpha wave activity during slow wave sleep. Originally considered a possible neurological contribution to FMS, whether the alpha-delta sleep anomaly is fundamental to the development of fibromyalgia syndrome, or results mainly from the pain experience of FMS patients remains unknown. No previous study has directly compared the sleep of FMS and other (non-FMS) patients experiencing similar levels of chronic pain and sleep dysfunction. The present study was designed to examine sleep macrostructure and microstructure in FMS patients, and evaluate the role of the alpha-delta sleep anomaly as either a possible contributor to fibromyalgia syndrome, or a likely consequence of pain experience. In order to explore these relationships, detailed sleep, activity and psychological profiles were compared in 3 groups: 1) FMS patients (n = 19); 2) osteoarthritis patients with sleep disturbance (n = 17); and non-clinical (normal healthy) adults (n = 10). In order to standardise diagnostic reliability and symptom chronicity, the FMS group was recruited from a single rheumatology facility immediately following diagnosis. Guided by a series of formal research questions, analyses compared sleep macrostructure (using American Academy of Sleep Medicine criteria), sleep microstructure (using spectral analysis), and a range of psychological variables (including pain experience, sleepiness, fatigue, depression, anxiety, perceived social support, health locus of control, pain catastrophizing and personality). The results indicated that the alpha-delta sleep anomaly is not unique to FMS, but appears to be a feature found in the sleep of normal healthy adults and (to a greater extent) those with FMS and osteoarthritis. The incidence of the anomaly was statistically similar in both clinical (FMS and osteoarthritis) groups, a pattern consistent of its being a secondary feature of pain, rather than a primary abnormality of FMS. Overall, the psychometric assessments of state and trait anxiety and depression better discriminated between the three groups than did the sleep variables. Nevertheless, on measures of sleep, perceived social support, health locus of control, and pain catastrophizing, FMS and osteoarthritis patients were not significantly different, though both clinical groups differed on these variables from healthy controls.

This Ph.D. dissertation processes and analyzes signals from the neonatal intensive care units (NICUs) for the study of maturity, systemic infection (sepsis) and the influence of immunization in the premature newborn. A special attention is payed to the electroencephalography and the breathing signal. The former is often contaminated by several sources of noise, thus methods based on the signals decomposition and optimal noise cancellation, adapted to the characteristics of the immature EEG, were proposed and evaluated objectively on real and simulated signals. By means of the EEG and delta burst analysis, detected automatically by a proposed classifier, infant's maturation and the effects of vaccination are studied. Concerning the second signal, breathing, non-linear and fractal methods are adapted to evaluate maturity and sepsis. A robustness study of estimation methods is also conducted, showing that the Hurst exponent, estimated on respiratory variability signals, is a good detector of infection.

A Real-Time Brain-Computer Interface is a technical system classifying increased or decreased brain activity in Real-Time between different body movements, actions performed by a person. Focus in this thesis will be on testing algorithms and settings, finding the initial time interval and how increased activity in the brain can be distinguished and satisfyingly classified. The objective is letting the system give an output somewhere within 250ms of a thought of an action, which will be faster than a persons reaction time.

Algorithms in the preprocessing were Blind Signal Separation and the Fast Fourier Transform. With different frequency and time interval settings the algorithms were tested on an offline Electroencephalographic data file based on the "Ten Twenty" Electrode Application System, classified using an Artificial Neural Network.

A satisfying time interval could be found between 125-250ms, but more research is needed to investigate that specific interval. A reduction in frequency resulted in a lack of samples in the sample window preventing the algorithms from working properly. A high frequency is therefore proposed to help keeping the sample window small in the time domain. Blind Signal Separation together with the Fast Fourier Transform had problems finding appropriate correlation using the Ten-Twenty Electrode Application System. Electrodes should be placed more selectively at the parietal lobe, in case of requiring motor responses.

Over the past decades, decision making under risk has garnered a great amount of attention both in the field of economics and psychology. Although state-dependent variabilities of risk taking are well-documented, little is known about the effects of a person’s preferred time of day, or chronotype, in risky decision making. Under circumstances of circadian mismatch (e.g., when an “early bird” makes decisions in the evening), research suggests that decision making may reflect a greater reliance on heuristics, such as using stereotypes in social judgments. However, the effects of circadian mismatch on heuristics in risky decision making are relatively unexplored. This paper looks into the effects of circadian mismatch on the reflection effect: a behavioral bias in financial decision making, wherein individuals are risk averse when facing potential gains, and risk seeking when facing potential losses. Participants will be randomly assigned to their circadian matched or circadian mismatched conditions and will play a series of financial gambling tasks with real monetary incentives. This study predicts that the reflection effect will be exacerbated in circadian mismatched individuals as compared to matched participants. Exploring such an effect could have real-world implications on decision making under risk by providing critical knowledge about the effects of time of day on our susceptibility to behavioral biases. It could therefore point to the existence of a more optimal time of day to engage in such critical decision making.

碩士
國立成功大學
醫學資訊研究所
106
Humans spend one-third of their lifetimes on sleep. One with good quality of sleep can improve his or her attention, memories and metabolism. However, not all of the humans can have good quality of sleep. For those people who have been plagued with sleep disorders should use a multi-channel polysomnography (PSG) to improve them at some selected hospitals. The PSG records whole-night sleep physiological signals and the measurement of brain activity (EEG), eye movements (EOG) and muscle activity (EMG) as the parameters for the experts to undertake the research for sleep stage scoring. The PSG has enormous and multiple biophysiological recording functions. During sleep recording, attaching a large number of electrodes leads to subjects has become one of the sleep disturbances to them, which often requires extra help from technicians. Compared with EEG, EOG uses electrodes placed around the eyes without blocked by thick hair on the forehead, where EEG signals are relatively accessible to be measured for sleep scoring. Therefore, we invented a set of eye-mask sensing device based on brain signals. In terms of hard drive device, we used low noise analog front-end (AFE) to design signals and harvest circuits in cooperation with a built-in Bluetooth of system on chip (SoC) for wired and wireless communication. In order to verify the convenience, accuracy, stability and applicability of this system, we conducted four types of experiments in this study. In Experiment 1, we collected 24 recording times of wearing eye-masks and 10 recording ones of setting up PSG to compare the two systems concerning their convenience. In Experiment 2, this system and PSG collected whole-night sleep recordings of both EEG and EOG signals on 11 healthy adults. This experiment attempts to prove that the signals are relevant and consistent with sleep staging scoring. In Experiment 3, for the purpose of daily uses, we collected recordings for 4 consecutive whole-night sleeps and 8 napping recordings to implement its stability. In Experiment 4, simultaneous eye movement detection algorithm was well applied to human-computer interaction games based on the structure of eye mask. According to the results of the three previous experiments, we suggested that this system and PSG have acquired 85% agreement with sleep scoring reaching up to the standard of clinical judgment at present. Furthermore, in comparison to PSG setting time of 47 minutes on average, this system enabled the subjects to spend only two minutes wearing it. Its wearable and convenient features were proven to be true on reducing time for its set up. For the research on human-computer interaction games, it was also carried out to detect eye movements in 0.377 seconds (standard deviation: 0.043) reaching up to 96% accuracy (standard deviation: 5.6). As discussed above, this system is expected to have significant effect on the measurement of sleep signals and human-computer interaction.

碩士
元智大學
機械工程學系
91
The control system of human central nervous system (CNS) is complex. Electroencephalogram (EEG) is one tool to explore the physiological function of brain. It is widely used to diagnose clinical CNS syndromes, such as epilepsy, brain tumor, Parkinson’s disease, etc. However, mostly commercial EEG instruments belong to multi-channel, high price and too big for portable usage. Hence, our thesis tries to design a low price, small volume and single channel of EEG device for easy portable usage. It can use for medical measurements .We used the fourier transform analysis for five volunteers at each time domain in the awake and sleep EEG signals. We found significant differences in sleep and awake waves form according to the volunteers’ sleep and awake statuses. The successful results are given us to have the confidence to test this EEG device in the clinical trials in the near future.

碩士
國立臺中科技大學
資訊管理系碩士班
103
Sleeping is a very important and crucial thing in people’s daily lives. Good sleeping quality has a huge effect on daily work or study performance, so we could not underrate its importance. Previous studies used sleeping stage or questionnaire assessment to evaluate sleeping quality. This study tried to propose three new overnight sleep EEG signals to evaluate sleeping quality. The signals were combined with data mining technique to develop a basic sleeping quality evaluation system. In 168 days, 36 adults were invited to attend our experiment. We obtained their sleeping data and evaluated their sleeping quality. Finally, the accuracy of this evaluation system had reached 87.5%. With the development of mobile technology, mobile devices have brought a lot of convenience to our live. People use many applications of mobile devices to do social behaviors, watch videos and play mobile games. Mobile devices do not only gradually affect our bedtime activities, but also has effect on our sleeping quality. This study explored the effect of bedtime activities on the sleeping quality and brainwave. Using a mobile device to do activities, such as gaming, watching videos, or chatting on the Internet before sleeping, would increase the proportion of high amplitude in each band(α, β, θ and δ), and then lead the quality of sleeping downward. Among these activities, watching the video had the hugest effect on sleeping quality. Furthermore, surfing the Internet and gaming before sleeping would affect the sleeping quality of the female participants. In addition, watching videos had more effect on males’ sleeping quality.

博士
長庚大學
電機工程學系
106
Sleep spindles are generated in the thalamus. Neurons project from thalamus to cortex, known as thalmocortical circuitry. Most of the electrophysiological studies of thalamocortical oscillations were in animals. With the advance of deep brain stimulation devices, we had chance to record sleep oscillations from anterior thalamic nucleus. Concomitant EEG and deep brain signals were recorded in long-term monitoring unit. The whole-night time-frequency coherence maps between EEG (C3, C4) and deep brain local field potential showed specific coherence patterns during non-rapid eye movement (NREM) sleep. Pooled coherence in the NREM stage was significant in slow, delta, theta and spindle frequency ranges. Spindle oscillations had high coherence only ipsilaterally, while slow oscillations had similar coherence either ipsilaterally or contralaterally. Lempel-Ziv complexities was measured and revealed that either deep brain signals or EEG signals revealed less complexity in deeper NREM sleep. Omega complexity over deep brain signals, EEG signals or combined also showed trends on decreasing complexity with deeper NREM sleep. Our studies suggest that thalamocortical circuitry oscillated in specific frequency ranges, and the complexity decreased as NREM sleep deepened spatiotemporally.

Tese de mestrado integrado, Engenharia Biomédica e Biofísica (Engenharia Clínica e Instrumentação Médica) Universidade de Lisboa, Faculdade de Ciências, 2017
Decoding arm movement trajectory from brain signals would allow motor impaired people to control an arm prosthetic. Studies show that we can estimate a vector that points in the direction of arm movements based on single motor neuron activity - the population vector. This type of recording requires the surgical insertion of electrodes in the cerebral cortex. Although such invasive recordings would offer high spatial resolution, noninvasive recording have the advantage of high temporal resolution and no need for surgery. Researchers have managed to decode movement properties from noninvasive brain signals with similar accuracy as from invasive recordings. But can we find a noninvasive analogous of the population vector, a vector that points in the direction of the arm movement? This was the motivation for this thesis. To approach this question we acquired EEG, EOG and kinematic data from 12 healthy subjects while they performed a rhythmic circular right arm movement. We analyzed the data in the time and frequency domains. In the time domain we explored mainly the data averaged over cycles. We found a pattern that looked as if the potentials in the scalp rotated with the arm. To better visualize this rotation, we fit one dipole per time-stamp in the averaged cycle data of each subject to describe the scalp’s potentials. The dipoles rotated along the cycle for all subjects, most of them in the same direction and plane of rotation, with exception for two subjects whose rotation was opposite and three subjects with a slightly different rotation plan. In the frequency domain, we used the Source Power Comodulation algorithm (SPoC), an algorithm that searches for components whose power correlates with a target variable, in our case, the arm kinematics. By applying this algorithm to 20-24 Hz band-pass filtered data, we found two components per subject, each calculated with different kinematic target variables. The results show components that when applied to the non band-pass filtered data, created signals whose power spectrum highly correlated with the given targets (the average of the absolute correlations being 85.5%). The physiological reason for both these phenomena is not entirely understood. To find the analogous of the population vector there is still a long way to go, and we hope this thesis was a first step towards it.
O cérebro controla direta ou indiretamente todas as ações do corpo humano, entre elas o nosso movimento. O movimento é uma capacidade fundamental ao ser humano e, por essa mesma razão, indivíduos que sofram de incapacidades motoras têm uma redução considerável da sua qualidade de vida. Uma interface cérebro-computador (mais conhecida pelo seu nome em inglês brain-computer interface (BCI)) é um sistema que permite o controlo de dispositivos externos usando sinais cerebrais. Esta tecnologia é particularmente interessante para pessoas com incapacidade motora uma vez que não necessita de input físico e poderia ser usada para controlar uma neuroprótese ou um braço robótico. Existem várias estratégias que possibilitam o controlo destes sistemas, mas para o controlo de uma prótese do braço seria preferível usar uma estratégia natural, que não implicasse uma aprendizagem exaustiva por parte do utilizador. Para esse fim, é necessário descodificar vários parâmetros motores de acordo com a intenção do utilizador, como por exemplo, a direção do braço. A possibilidade de um dia conseguir descodificar sinais cerebrais para o controlo de dispositivos externos já começa a ganhar forma, mas ainda não é possível a um nível suficientemente eficaz. Usando métodos invasivos de aquisição de sinais cerebrais que requerem cirurgia para implantar elétrodos no córtex cerebral, Georgopoulos et al. conseguiram distinguir entre movimentos direcionais (em 8 direções num plano horizontal) em macacos. Nessas experiências criou o conceito de vetor de população (population vector) que é um vetor calculado a partir da atividade de neurónios motores que tem a particularidade de apontar na direção do movimento executado. Já no campo dos métodos de aquisição não-invasivos podemos destacar o eletroencefalograma (EEG) e o magnetoencefalograma (MEG) que adquirem sinais elétricos e magnéticos (respetivamente) com sensores colocados fora do crânio. Vários investigadores usaram estes métodos de aquisição para descodificar sinais cerebrais durante tarefas de movimento direcionais usando regressões lineares em sinais de baixa frequência, e modulações em frequência para sinais na gama dos 50-90 Hz (banda de frequência ϒ) e em frequência mais baixas para os 10-30 Hz (bandas de frequência α e β). Algo que ainda não foi estudado é a possibilidade de encontrar um análogo ao vetor população usando métodos não-invasivos. Este não teria os mesmos princípios do vetor de Georgopoulos, uma vez que nos é impossível inferir a atividade de neurónios singulares em métodos não-invasivos, mas teria o mesmo objetivo: apontar na direção do movimento executado. Para explorar este conceito realizámos aquisição de dados EEG, eletrooculograma (EOG) e dados cinéticos do braço direito de 12 sujeitos saudáveis, enquanto estes executavam um movimento rítmico, circular, no sentido dos ponteiros do relógio num plano vertical à sua frente. Durante a aquisição, os sujeitos focaram o seu olhar numa cruz mostrada através de um monitor colocado a sua frente, de forma a minimizar os movimentos oculares. Adicionalmente, uma divisória foi colocada perto do lado direito da face de cada sujeito impedindo os mesmos de observarem o seu braço enquanto realizavam o movimento requisitado, não obtendo assim qualquer feedback visual do seu membro superior. Os dados cinéticos foram adquiridos com um sensor Kinect para a Xbox 360 que ao longo da experiência localizou as junções do braço direito dos sujeitos. Os dados cinéticos foram filtrados com um passa-banda 0.3-0.8 Hz e, ao longo dos ciclos do braço, os pontos extremos do braço (i.e., os máximos e mínimos nas coordenadas vertical e horizontal) foram anotados nos dados para possibilitar a associação dos sinais cerebrais com a trajetória do braço em cada ciclo. Para cada sujeito os canais EEG ruidosos foram interpolados, os dados foram referenciados à média comum de todos os canais, e os sinais foram filtrados numa banda de frequência 0.25-100 Hz e com um filtro tapa banda nos 50 e nos 100 Hz, este último para rejeitar o ruído de fundo. Os sinais de EEG e EOG foram separados em épocas conforme a posição do braço, sendo que cada época passou então a consistir num ciclo do braço completo que começa no ponto mais alto da coordenada vertical. Cada época foi depois temporalmente distorcida para que todas tivessem a mesma duração. As épocas com artefactos foram rejeitadas da análise usando métodos automáticos de rejeição. Independent Component Analysis (ICA) foi utilizada para identificar e posteriormente rejeitar componentes independentes referentes a movimentos musculares e oculares. Por fim, os dados foram explorados em ambos os domínios de tempo e frequência. No domínio do tempo, estudámos mais especificamente a média das épocas de EEG e EOG durante os ciclos do braço. Uma vez que sinais não-invasivos são muito sujeitos a ruído, a média elimina artefactos singulares e acentua os sinais que aparecem constantemente nos dados. Os sinais do ciclo médio mostraram um padrão interessante para todos os sujeitos; um comportamento rotacional ao longo da rotação do braço direito. Para acompanhar a rotação dos potenciais, procurámos por um dipolo que descrevesse a distribuição topográfica a cada ponto do tempo. A rotação dos potenciais do EEG ao longo do ciclo médio foram verificados com a rotação da direção do dipolo ao longo do ciclo. A grande maioria dos sujeitos obteve um dipolo a rodar no mesmo sentido no mesmo plano (segundo a regra da mão direita, com um vetor de rotação a apontar para a zona frontal esquerda do cérebro). Cinco sujeitos foram a exceção, 2 desses cujo dipolo rodava no sentido contrário, e os restantes 3 sujeitos cujo dipolo rodava no mesmo sentido, mas num plano ligeiramente diferente. Em todos os sujeitos o dipolo ajustado rodava, de forma relativamente uniforme. No domínio da frequência, estudámos em particular a banda de frequência dos 20 aos 24 Hz. Escolheuse esta banda de frequência pois demonstrou os resultados mais interessantes e já tinha sido utilizada em estudos prévios. Usámos um algoritmo chamado SPoC (Source Power Comodulation) que encontra componentes de atividade cerebral cuja amplitude em frequência correlacione com uma variável alvo. Como variável alvo usámos os dados cinéticos do braço direito, e como input os dados cerebrais filtrados por um filtro passa-banda (20-24 Hz). Os resultados traduziram-se numa série de componentes cuja amplitude correlacionava ou anti-correlacionava com o movimento do braço, muitas delas com projecções topográficas consistentes com as áreas cerebrais motoras. Encontraram-se algumas semelhanças entre os padrões de ativação das componentes do SPoC dos vários sujeitos, ainda que os resultados variassem entre cada um. Ao projetar as componentes aos dados não-filtrados pelo passa-banda, verificamos que as modelações em frequência de facto correlacionam com as variáveis-alvo como esperado, com uma média da norma das correlações de todos os sujeitos a 85,5%. No domínio temporal, ainda que recorrendo à média de todos os ciclos (épocas), este é o primeiro estudo que demonstra de forma não-invasiva, a existência de um dipolo com comportamento rotacional ao longo da rotação do braço. Para o seu uso em tecnologias de BCI, é necessário encontrar o mesmo fenómeno em épocas únicas, tornando possível uma classificação em single-trial e em tempo real. No que toca aos resultados no domínio da frequência, a procura por componentes cuja fonte poderia estar envolvida na criação do movimento circular foi também bem-sucedida. Este estudo abriu portas para uma série de investigações futuras. Para trabalhos posteriores destaco a necessidade de uma análise estatística, de usar mais do que um dipolo para descrever a distribuição de potenciais no domínio temporal, de explorar os dados em cada movimento e não apenas a sua média, e de explorar paradigmas semelhantes durante o movimento do braço esquerdo. Os resultados desta tese serviram, portanto, como primeiro passo na direção de encontrar o análogo não-invasivo do vetor de população.

碩士
南台科技大學
電機工程系
98
In clinical studies, EEG is frequently used to diagnose psychological diseases. When people are awake, visual evoked potential induced by visual stimulation is commonly used to assess whether there are lesions on the visual nervous system, such as optic neuritis. While during sleep, the sleep EEG stage is often used to assess whether patients has sleep disorders, such as insomnia, apnea syndrome, atrophy etc. However, integrated EEG analysis software is not yet available on the market. Therefore, the focus of the study is to develop such on integrated system that can provide a broad EEG-based academic research. This study has established a LabVIEW-based EEG analysis system. The system is divided into two subsystems: (1) Visual evoked potential analysis system: its internal functions include the realization of EEG through spectral analysis, filter selection and the use of independent component analysis (ICA). Moreover, wavelet analysis is utilized to carry out noise filtering and ultimately the ERP image and pattern reversal-visual evoked potential (PR-VEP) as its results; (2) Sleep EEG staging sub-system: The system a single channel EEG from a PSG device or from the EEG module developed by the team. Signal correction is done by using wavelet to remove noise signals. Afterwards, data can be processed using approximate entropy or sample entropy before passing through a rule-based classification to determine the different sleep stages. In order to verify the feasibility of the visual evoked potential analysis system, an auricular electrical stimulation was designed to perform experimental analysis. EEG data were collected from a commercial device, NeuroScan, and then the analysis of evoked potentials was done using the EEG software, EEGLAB and the system’s software. Preliminary results showed that the extracted latency of the visual evoked potential characteristics with stimulus frequencies of 1 Hz, 3Hz and 5 Hz produced by EEGLAB has statistical significance (p<0.05). As a corollary, some differences may occur between EEGLAB and the developed program due to some internal setup differences. Moreover, for the verification of the sleep stage classification subsystem, data were taken from the MIT-BIH database. First, wavelet transform was used to effectively remove noise. Second, the approximate entropy and sample entropy was used to compute the sleep stage to determine the characteristic changes of the EEG signals. It showed that wake has the largest value and a gradual decrease as sleep enters into stage 1 and 2. When sleep stage 3 and 4 was reached the lowest entropy values were computed. But as sleep enters into REM the value falls between stage 1 and stage 2. Currently REM value is situated around stage 1, while the values of stage 3 and 4 are so close that the two can be combined and renamed as slow wave sleep (SWS). However sleep staging using this method still has room for improvement. As of the moment preliminary results when compared with the expert diagnosis of MIT-BIH database reached an accuracy of 62% but REM cannot be determined clearly. In the future, a more advanced classification system would be applied to do the classification.

碩士
國立政治大學
心理學研究所
99
Introduction：Insomnia is a common healthy complain. The neurocognitive perspective of hyperarousal model of insomnia, as proposed by Perlis(1997), hypothesized that the sleep difficulties in insomniacs may result from enhanced information processing around sleep onset and during sleep. Supporting evidences were primarily from the findings that insomnia patients have increased high frequency EEG activity and decreased low frequency EEG activity during sleep, indicating insomniacs in general have higher physical arousal and lower sleep homeostasis. This study further aims to explore arousal level and sleep homeostasis during the period of sleep onset by comparing the level and change of EEG spectrum in primary insomnia patients and normal control subjects during the process of sleep onset. Methods：30 patients with primary insomnia (10 men, 20women, mean age of 36.7years) and 25 normal sleepers (8 men, 17women, mean age of 34.8years) underwent one night of PSG recording in a sleep laboratory to screening sleep-related breathing disorders and sleep-related movement disorders. They also completed the Pre-sleep Arousal Scale (PSAS) before bedtime. EEG spectrum analyses were conducted for the EEG data collected during the 5 minutes prior to sleep onset and the 15 minutes after. Results：Subjective ratings of both pre-sleep cognitive and somatic arousal were significantly higher in insomnia group (F = 23.950, p &lt; .001; F = 64.235, p &lt; .001) than control group. More WASO (F = 5.510, p = .023), less time and percentage of stage 2 sleep (F = 7.088, p = .010; F = 32.616, p &lt; .001), less percentage of REM sleep (F = 4.810, p = .033), and poor sleep efficiency (F = 8.685, p = .005) were showed in PSG. The EEG spectrum during sleep-onset period showed that insomniacs had higher alpha power in the sleep-wake transition, lower delta power after falling asleep, and higher theta and beta power during sleep-onset period. In terms of the slope of EEG specrtrum change during the period of sleep onset, insomniacs had slower change than normal sleepers in increasing of sleep homeostasis and decreasing of physical arousal. In addition, the correlations between PSAS score and EEG power, cognitive arousal and delta power after falling asleep and theta power in sleep-onset process showed significant positive correlation. Alpha power in the later part of sleep-onset period and beta power around sleep-wake trainsition, on the other hand, showed negative correlations with cognitive arousal. Physcial arousal only showed positive correlation to theta power in sleep-wake trainsition. Conclusions：Patients with primary insomnia showed significantly less and slower increase in sleep homeostatic drive as well as less and slower decrease in EEG arousal during sleep-onset period. Although EEG arousal did showed gradually decreased by time, it still maintianed higher than normal sleepers. Sleep homeostasis did also increase, but may be interfered by the hyperarousal. This may explain the complaints in insomnia patients of difficulty falling asleep, difficulty maintaining sleep, and light sleep.

碩士
國立陽明大學
腦科學研究所
105
Background: Thermal environment affect the sleep quality; however, there lacks the measurements for assessment of thermal environment during sleep. The thermal comfort during sleep still needs other objective parameters. Furthermore, it is also important to provide suitable environmental temperature for different sleep stages. Therefore, we want to investigate the influences of different environmental temperature on sleep-related physiological signals. Methods: This study is divided into two parts. Firstly, to measure the changes of autonomic nerves system (ANS) functions, electroencephalogram (EEG), electrocardiogram (EKG), and subjective feelings among different temperature during awake. Secondly, to explore the effects of different temperature on objective sleep quality. Analysis: The EEG, EKG data are received by miniature polysomnography, made by K&Y lab, and using FFT to analyze the ANS function and brain activity. Result: From the first part of experiment, we find the most comfortable temperature is 26 °C. When subjects in 26 °C, the Theta Power % is higher and Beta Power % is lower than other temperature. The RR variability in 22 °C, 24 °C is significant higher than in 26 °C and 28 °C. There is negative correlation between RR variability and temperature. (r = -0.309, P<0.01). In the second part of experiment, male subjects have significant longer slow wave sleep. But there is no significant different between subjective sleep quality questionnaire and the temperature. Conclusion: Different temperature will affect subjective feeling, physiological parameters, and sleep. According to the relationship between physiological parameters and subject questionnaire, we may build up the assessment of thermal comfort indices.

碩士
國立成功大學
資訊工程學系
102
Sleep occupy more than one-third of human life, have a good sleep for the quality of life has a significant help. But not everyone has a good quality of sleep, many people have been plagued by sleep-related disorders. Therefore, the clinical use Polysomnog- raphy (PSG) to recording sleep physiological signals. The collection of physiological signals will be manual scoring by expert for diagnosis. Since manual scoring is a very subjective and time-consuming work, so there are many sleep automatic scoring methods have been proposed. Although the agreement of these methods on a good performance, but only provides a final interpretation of the results. Experts can not know the basis of automatic scoring methods, so experts need to re-scoring when the scoring result not to be believed, in this way automatic scoring method can not achieve its designed purpose, to reduce interpretation time. Therefore, this study proposes a human-computer cooperation scoring system which based on sleep physiological signals, to providing reliability as reference. Advantages of this system is the scoring results is divided into two parts, high reliability and low reliability. Scorer can skip the high reliability epochs to reduce scoring time, and let experts believe the overall scoring re- sults. The system has been testing by two scorer, the average agreement of full manual scoring and work with cooperation systems can reach 88.47%, kappa coefficient was 0.82, and can reduce 56.2% scoring time. We hope this human-computer cooperation scoring system can practical on clinical applications, providing a more reliable scoring results, in addition to reduce scoring time.

碩士
國立臺北科技大學
機電整合研究所
103
The research is a wearable device system with sensors wear on human wrist which can receive the information of skin conductance, skin temperature, heart rate and accelerometer signal from sensors used in judgment of the Rapid Eye Movement (REM) sleep stage. Furthermore, we also study the relationship between the skin conductance variability and the environmental comfort. Then we try to qualify the skin conductance in different environmental comfort. We wish to use this quantization result in automatic air-condition controlling to reach the adjustment of environmental comfort and improve sleep quality to solve insomnia due to the uncomfortable environment in the future. In addition, our wearable device can transmit human information to BLE embedded lights mesh which are built by laboratory. Bluetooth devices embedded on lights mesh will transmit the information to Wi-Fi modules with shortest path after that users can use mobile devices to get the real-time information from cloud to achieve the real-time monitoring purpose further or to build a private database.

abstract: The human brain controls a person's actions and reactions. In this study, the main objective is to quantify reaction time towards a change of visual event and figuring out the inherent relationship between response time and corresponding brain activities. Furthermore, which parts of the human brain are responsible for the reaction time is also of interest. As electroencephalogram (EEG) signals are proportional to the change of brain functionalities with time, EEG signals from different locations of the brain are used as indicators of brain activities. As the different channels are from different parts of our brain, identifying most relevant channels can provide the idea of responsible brain locations. In this study, response time is estimated using EEG signal features from time, frequency and time-frequency domain. Regression-based estimation using the full data-set results in RMSE (Root Mean Square Error) of 99.5 milliseconds and a correlation value of 0.57. However, the addition of non-EEG features with the existing features gives RMSE of 101.7 ms and a correlation value of 0.58. Using the same analysis with a custom data-set provides RMSE of 135.7 milliseconds and a correlation value of 0.69. Classification-based estimation provides 79% & 72% of accuracy for binary and 3-class classication respectively. Classification of extremes (high-low) results in 95% of accuracy. Combining recursive feature elimination, tree-based feature importance, and mutual feature information method, important channels, and features are isolated based on the best result. As human response time is not solely dependent on brain activities, it requires additional information about the subject to improve the reaction time estimation.
Dissertation/Thesis
Masters Thesis Electrical Engineering 2019

碩士
國立交通大學
多媒體工程研究所
102
Human cognitive limitations are becoming a crucial issue in human computer interaction nowadays and thus an important determinant of overall human performance. This thesis purpose to introduce two methods which applying EEG technology help us to improve and evaluate human computer interaction. First, we suggested a new approach to detect users’ attention state in chapter 2. Attention monitoring is particularly important for many HCI applications. How to automatically determine a users visual attention state is challenging since attention involves many complex and internal human cognitive functions. Behavioral observations, such as eye gaze or response to external stimuli, can provide some clues for users visual attention state; however, the users cognitive state cannot be easily known. We explored the feasibility of designing an attention monitoring system that can detect if our brain sees a visual stimulus consciously. Second, we used an EEG-based approach to assist usability test with audio notifications in chapter 3. Audio notifications have become an important way to prompt users. Several studies have been proposed to evaluate audio notifications, but they rarely considered user workload and environmental impact in the same time. We developed an EEG-based approach to evaluate audio notifications by measuring subjects’ auditory perceptual response (mismatch negativity) and attention status (P3a). We demonstrated this approach by two experiments, in which auditory icons were evaluated under different workload and environments. According to the experiment results, the perceptual effects of audio notifications could be measured objectively. Based on these technologies, we could directly detect users’ cognitive changing when they interact with our design. Therefore, this thesis used EEG device to detect and analysis variation of human cognitive state from visual and auditory perception respectively. We hope to provide new feasible approaches and view to HCI field.