Our Talented People
李哲宇 (Autumn 2018); email@example.com
邱載峰 (Autumn 2018); firstname.lastname@example.org
廖堉翔 (Autumn 2018; 五年一貫); email@example.com
李偉銘 (交通大學碩士班); firstname.lastname@example.org
Terence Liu (交通大學博士班); email@example.com
曹澤宏, 交通大學博士班/University of Technology Sydney, 2017
Developing a Migraine Attack Prediction System using Resting-state EEG (基於靜息狀態腦波之偏頭痛預測系統)
Migraine is a common episodic neurological disorder with complex pathophysiology characterised by recurrent headaches during a period like one month. Only small group of migraine patients (13-31%) experienced transient neurological symptoms that are most frequently visual aura prior to headache onset, yet a majority of patients were migraine without aura (MO) that do not possess the premonitory symptoms. This study explored neurophysiological evidence of the resting-state electroencephalogram (EEG) power, coherence and entropy to support the cortical signals correlate of different migraine phases, and then develop an EEG-based system for predicting migraine attacks. First, we investigated EEG devices, pre-processing and artefact removal methods, and feature extraction technologies, including power, coherence and entropy analysis. Next, we discovered cyclic EEG dynamics of migraine on the cross-sectional basis. The results indicated that EEG power spectral and coherence were significantly increased in the pre-ictal group, relative to EEG data obtained from the inter-ictal group. Inter-ictal patients had decreased EEG power and connectivity relative to healthy controls, which were “normalised” in the pre-ictal patients. Furthermore, on the basis of longitudinal design, we estimated brain dynamics before migraine attacks using a wearable EEG device. The results showed the EEG entropy of individual patients in the pre-ictal phase, resembling normal control subjects, was significantly higher than that in their inter-ictal phase in prefrontal area. That is, the entropy measures identified enhancement or “normalisation” of frontal EEG complexity in pre-ictal phase. Finally, based on these neuroscience discovery of inter- and pre- ictal EEG entropy in individuals, this study proposed a support vector machine (SVM) based system with 76% accuracy to predict migraine attacks. The prediction system characterises the EEG entropy of single (prefrontal) area favoured the application of brain-computer interface in migraine.
王瀚君, 交通大學碩士班, 2017
Detecting Human Brain State Drift Based on Stationary Subspace Analysis (基於穩態子空間偵測大腦狀態變換)
Brain state drift, such as from alertness to drowsiness, leads to numerous miserable incidents. In order to prevent the tragedy resulting from human state changes happens in the future, detecting the state drift becomes a crucial issue. Previous studies mentioned that the state shift may relates to nonstationarities in electroencephalography (EEG). Therefore, we analyze the nonstationarities of EEG and propose an indicator, the number of stationary components, which is decomposed by a novel algorithm, Stationary Subspace Analysis (SSA), to monitor the human state drift. Moreover, to understand the full spectrum of stationary dynamics changing with brain state, we assessed two datasets: one is lane-keeping task (LKT) dataset, the other is sleep dataset. In LKT dataset, the behavioral results showed that human in alert state has more stationary components than that in drowsy state. Significant negative correlation between driving reaction time (RT) and the number of stationary components was found. Furthermore, significant negative correlation between beta band power of both stationary components and non-stationary components and RT was found. In sleep dataset, human in wakefulness state has more stationary components than that in N1, N2, and N3 stage. Nevertheless, the number of stationary components rose in REM stage. Nonetheless, there is a significant negative correlation between the number of stationary components and the sleep stage before first N3 stage occurred. Apart from that, significant positive correlation was found after people enter deep sleep until waking up. In conclusion, the number of stationary components can be utilized to detect the brain state drift. Such findings could largely reduce the rate of traffic incidents caused by brain state drift.
洪鈺嘉, 交通大學碩士班, 2017
An EEG-Based Driving Performance Prediction System using 3D/4D Convolutional Neural Network (使用三維/四維卷積神經網路的腦電波系統用於預測駕駛表現)
Motor vehicle accidents have been the leading cause of fatalities worldwide. Most of those accidents are due to mistaken operation and behavioral lapses and majorly are caused by drowsiness and fatigue. In other words, predicting the cognitive state of drivers or even monitoring the reaction time (RT) while driving is a potential research. To predict activity of human brain, electroencephalogram (EEG) has been proved be an indicator of monitoring the state of human brain and human behavior. In EEG signal analysis, the most adopted method is machine learning since the strength of auto feature extraction. On the other hand, deep leaning algorithm has been shown improving the accuracy of machine learning in many fields. Some works have adopted deep learning algorithm in EEG signals analysis already. However, the analysis of EEG signals can be extremely challenge, since the signals is relatively complicated in signal processing. In traditional deep learning method, the essential information of EEG signals cannot be considered into analysis efficiently. For example, temporal information and the spatial information of EEG channels are indispensable in EEG signal analysis but cannot be extracted well in current methods. In order to solve this problem, this study applied a proposed convolutional neural network based (CNN-based) algorithm, 3D convolutional neural network (3D CNN) to solve the temporal information. In other words, by applying 3D CNN on EEG signal analysis, the module can extract temporal information, frequency information in EGG channels. Furthermore, the study also proposed a novel deep learning method 4D convolutional neural network (4D CNN), which is also a CNN-based algorithm. In terms of feature extraction of EEG signals, 4D CNN is able to combined different attributes, frequency, temporal information and spatial information of EEG channels together by performing four convolutions in feature extraction. By doing so the model can map EEG-RT relationship precisely in the cognitive state monitoring system. The contribution of this work is improving deep learning method in the analysis of EEG signals according to the knowledge of brain researches.
曾俊穎, 交通大學碩士班, 2017
EEG Information Transfer Changes on Fatigue Drivers (疲勞駕駛者之腦電波資訊流變化)
Traffic fatalities are one of the leading causes of death in the world. The motor vehicle crash was a serious problem in the past decade. Previous research has suggested that driving behavior and performance played an important role in driving safety. Therefore, a comprehensive understanding of the neurophysiological makers of declining driving performance during driving is significant. Several studies have focused on the change of spectral power in specific brain regions during simulated driving. Another research suggested that the change of brain connectivity in a specific cortico-cortical pathway may also be a sensitive neurophysiological signature for changes in alertness. Additionally, previous studies have stated that the fatigue would change the relationship between the EEG spectral power and behavior during driving. Based on the research above, we wanted to identify the effect of realistic fatigue on brain connectivity-behavior relationship during driving. Thus, in the study, we used the actigraphy device to access the fatigue level of 17 subjects. When the fatigue met the criterion levels, subjects would be asked to participate in the lane keep task. The EEG data were divided into three groups based on different levels of fatigue. We calculated the Transfer entropy of the EEG data to get the effective connectivity of each subject. The result showed that inverted-U shape change of connectivity was founded from high performance to poor performance only in low- and median-fatigue groups. The result demonstrated that different kinds of decreasing shape of connectivity magnitude from high performance to poor performance appeared in different groups. We observed that there was shift inverted-U shape in low- and median-fatigue groups. Additionally, we observed the connectivity difference between low- and high-fatigue groups. The result showed that the magnitude of connectivity decreased at the frontal region and increased at the occipital region form the low- to high-fatigue groups. In a nutshell, these results showed that different levels of fatigue would affect the relationship between brain connectivity and the behavior during driving.
連培中, 交通大學碩士班, 2017
Data Stream Mining Technology for ECG Signals of Chronic Pain: Real-Time Tracking and Clinical Correlation (慢性疼痛心電圖生理訊號資料串流深勘技術：即時追踪與疼痛臨床關聯)
Evaluating and tracking the progress of treatment for chronic pain is challenging because pain is a subjective experience and can be measured only by self-report. Electrocardiography (ECG) has been proven to be a promising source of physiological biomarkers for chronic pain. Previous studies had demonstrated that heart rate variability (HRV) could be associated with different types of pain and also pain perception. This study aims to identify the relationship between HRV indices and chronic pain through collecting resting ECG data and subjective pain severity from patients with chronic migraine and fibromyalgia before and after treatments. In addition, resting ECG data from healthy controls were also collected for comparison. The results derived from time, frequency, and non-linear analyses showed that the HRV of chronic patients were generally lower than that of healthy control subjects. Besides, the HRV of the chronic pain patients in the responder group significantly increased after the medical treatment, indicating that a useful biomarker of the treatment efficacy. Among 10 HRV indices, the non-linear Poincaré plot analysis is a promising HRV indices in monitoring pain severity as well as determining treatment efficacy. Finally, a data stream mining platform was developed for real-time streaming and analyzing of multimodal data. This platform is presented such that they can be used as an aid for biofeedback treatment of chronic pain in the future.
陳柏蒼, 交通大學碩士班, 2016
Dynamically Weighted Ensemble-based Prediction System for Adaptively Modeling Driver Reaction Time (動態權重調整之整合系統應用於駕駛者反應時間預測)
Motor vehicle crashes are the leading cause of fatalities in the US. Most of these accidents are caused by human mistakes and behavioral lapses, especially when the driver is drowsy, fatigued, or inattentive. Clearly, predicting a driver’s cognitive state, or more specifically, modeling a driver’s reaction time (RT) in response to the appearance of a potential hazard warrants urgent research. Recently, the electric field that is generated by the activity of the brain, monitored by an electroencephalogram (EEG), has been proved to be a robust physiological indicator of human behavior. However, mapping the human brain can be extremely challenging, especially owing to the variability in human beings over time, both within and among individuals. Factors such as fatigue, inattention and stress which can induce homeostatic changes in the brain, which affect the observed relationship between brain dynamics and behavioral performance, and thus make the existing systems for predicting RT difficult to generalize. To solve this problem, an ensemble-based weighted prediction system is presented herein. This system comprises a set of prediction sub-models that are individually trained using groups of data with similar EEG-RT relationships. The prediction outcomes of the sub-models are predicted by the weights that are derived from the EEG theta power and coherence, whose changes were found to be indicators of variations in the EEG-RT relationship, to obtain a final prediction. The results thus obtained reveal that the proposed system with a time-varying adaptive weighting mechanism significantly outperforms conventional systems in modeling a driver’s RT. The adaptive design of the proposed system demonstrated its feasibility in coping with the variability in the brain-behavior relationship. The contribution of this work is that simple EEG-based adaptive methods are used in combination with an ensemble scheme to increase significantly system performance.
謝濟遠, 交通大學碩士班, 2015
A Longitudinal Study of the Effect of Fatigue on Brain-Behavior Relationships in Driving (疲勞影響駕車時大腦與行為關係之縱貫性研究)
Drowsiness can impair task performance and increase behavioral lapses during driving, leading to becoming one of the main causes of fatal car crashes in the past decade. Previous studies have systematically established the brain-behavior relationship through the electroencephalography (EEG) signals recorded from human subjects when they performed a lane-keeping task in a simulated driving environment. Fatigue is the main cause of drowsy driving. However, the effect of fatigue on brain-behavior relationships is unclear. Thus, in this study, we developed the Daily Sampling System integrated with actigraphy and questionnaires to longitudinally assessed and tracked objective and subjective fatigue level of 17 subjects, respectively. When the fatigue met the criterion levels, subjects would be asked to participate in the driving experiment. The behavioral results showed that the reaction time (RT) in response to the deviation event increased with increased fatigue. The pre-stimulus brain activity showed that EEG theta and alpha powers of most of the brain regions observed in low- and median-fatigue groups increased as the RT increased. In the high-fatigue group, the theta power of posterior brain regions dramatically increased with the increased RT as compared with those in low- and median-fatigue groups. Additionally, the alpha power of the occipital regions showed an inverted U-shaped change which was observed only in the high-fatigue group. Taken together, fatigue significantly affects the brain-behavior relationship. Such findings could have major implication for understanding fatigue in drowsy driving and its model.
蔡岑, 交通大學碩士班, 2013
Effect of time pressure on inhibitory brain control for emergency driving (時間壓力對於緊急駕駛行為之大腦抑制控制之影響)
How to deal with the upcoming emergency situations is a key to avoid car accidents. Previous study (Chen, 2013) used brain imaging to reveal that the efficiency of inhibition function is responsible for copying such situations. However, other factors, such as stress, on driving inhibition are still unknown. Hence, in this study, we aim to get an insight into brain activities of emergency management in stress conditions. To investigate driver’s brain responses to inhibition function, a modified stop-signal driving task was implemented in a virtual-reality driving environment. The electroencephalography (EEG) was recorded from 16 subjects as they performed the experimental tasks under normal (without time pressure) and stress (with time pressure) conditions. Given a fixed road distance, each subject was instructed to arrive at the finishing line within a limited time under the stress condition. In signal processing, independent component analysis (ICA) and event-related spectral perturbation (ERSP) analysis were applied to investigate the spectral dynamics of independent brain processes. The behavioral results showed that the stop-signal reaction time (SSRT) was shorter under the stress condition than that under the normal condition. This result indicated that the stress could help to improve the efficiency of inhibition ability. The ERSP results showed that the augmentation of delta (1-3 Hz) and theta (4-7 Hz) powers in frontal and central areas are related to the inhibition mechanism. There is no statistically significant difference between two conditions. However, beta (13-30 Hz) and gamma (30-50 Hz) powers in frontal and central areas increased only in the stress condition. The beta and gamma powers of the central area under the stress condition were significantly higher than those under the normal condition. Because the gamma band is thought to reflect the top down modulation, the time pressure could possibly improve the driving inhibition efficiency by the proactive control which prepares to stop before the signal onset.
莊維彧, 交通大學碩士班, 2013
Hemoglobin and EEG correlates of fatigue during driving (駕駛員之大腦血氧濃度及腦電波與疲勞的相關性)
Mental fatigue is an important issue because it can be easily induced by prolonged task while not being noticed. Therefore, in this study, we will focus on the influence of fatigue during simulated driving on brain dynamics and hope to explore the relationship between electrical and hemodynamic features for further development in safety oriented supporting system. The event-related lane-departure driving task was implemented for 16 subjects with a combined electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) approach. In tonic analysis, we found that in theta (4-7 Hz), alpha (8-12 Hz) and beta (13-30 Hz) power strongly correlated as reaction time increased; the oxygenated hemoglobin (HbO) concentration increased as fatigue level rose but decreased when failed to perform the task well. Besides, phasic EEG demonstrated the event-related desynchronization (ERD) regarding deviation onset in the four power bands; HbO activation in phasic phase was declined as performance became worse. Furthermore, the negative correlations between tonic EEG delta, alpha power and fNIRS HbO oscillations were the most significant, thus suggesting that HbO might also be considered related to subjects’ mental fatigue level comparative to EEG alpha power.
陳諺玄, 交通大學碩士班, 2012
Brain Dynamics of Inhibition Process In Emergency Driving (緊急駕駛行為下之大腦抑制過程之運作)
Emergency situations while driving is difficult for driver to deal with, and it might lead to a critical damage for drivers. In this study, we aim to obtain a better and precise understanding of brain activities of inhibition mechanism in emergencies situations, which is of great benefit for devolvement of driving assistant system. Two virtual realistic driving experiments based on stop signal paradigm were performed for 15 subjects to study Electroencephalography (EEG) brain dynamics. In addition, Independent component analysis (ICA) and event-related spectral perturbation (ERSP) analysis were applied for investigating power changes in emergency driving. We found that theta (4-7Hz) and delta (1-3Hz) power increases in Frontal and Central areas reflect inhibition process. In previous inhibition studies, beta power increase in Frontal area was also related to inhibition response. However, our results suggest that this power change in Frontal area might not be a reliable EEG characteristic of inhibition in emergency driving. Furthermore, by comparing successful and unsuccessful inhibition, this study found that delta band power activities in Central area could be an EEG index for implementing prediction systems.
王潔明, 交通大學碩士班, 2011
Study of Effective Connectivity in Human Brain Network under Drivers’ Different Arousal Levels (探討駕駛人員腦內網路之有效連結在不同意識狀態下的改變)
Drowsy driving is one of the major factors leading to traffic accidents, especially occurring in a monotonous environment, the night-time driving, or after long-term driving. To avoid the occurrence of drowsy driving, a considerable number of studies attempted to develop an in-vehicle protocol via monitoring the electroencephalogram (EEG) features of drowsiness. One of the promising measures to evaluate the cognitive state is the change of EEG power spectra. However, most of previous literatures focused on the neurocognitive characteristics on separate brain regions, the human brain network in response to the cognitive transition from alertness to drowsiness is yet poorly understood. To address this issue, this study applied independent component analysis and partial directed coherence to show the change of effective connectivity between distributed brain regions under different vigilance levels, including alertness, transition, drowsiness, and abrupt-awake, during the simulated driving. The results of alpha coupling showed that the extrastriate cortex sent a causal outflow to the anterior region and received a causal inflow from the posterior region while being alert, compared to being drowsy. Regarding the transition state, the anterior region played a major source to affect the rest of the brain region with a cross-frequency coupling, and the connectivity magnitude had a relatively large causality, compared to other vigilance levels. Most of causal magnitudes declined as subjects progressed into a drowsy state. Interestingly, the subjects enabled a short reaction time in response to traffic events when they abruptly awakening from the drowsy state, however, the causal magnitude climbed to the level as the transition state, rather than the alert state.