- Email: [email protected]
High-performance computing system for electroencephalographic data analysis
Abstracts / International Journal of Psychophysiology 77 (2010) 288–342
stuttering is not an insulated (only peripheral) speech disorder, but it is part of a syndrome of particular mnestic, neurodynamic and motor defects revealing a dysfunction of deep posterior frontal and middle structures of the brain (functional blocks I and III according to Luria), and comprehensive neuropsychological and psychotherapeutic treatment can reduce these problems. 2. Psychophysiological results. Discrepancy in the potential of speech readiness in normally speaking and stuttering subjects is related to the pathological overactivation of the limbic system and frontal cortex and pathological deactivation of the temporal cortex, brainstem and basal nuclei. Also, the coordination of hemispheres in the process of preparation for speech is very weak in stutterers. Treatment of stuttering changes frontal activity, reduces limbic activity and reactivates the brainstem and temporal cortex. Generally, the longitudinal study revealed the changes of many psychophysiological parameters after speech rehabilitation including reintegration of activated and deactivated areas into a more normal pattern of brain activity and improvement of inter- and intrahemispheric interaction. Conclusion: Collected data allow us to update our knowledge about general neurocognitive mechanisms of stuttering and, particularly, about mechanisms of readiness for speech in a specific kind of distributed functional verbal system, such as stuttering.
311
excitation) and excitation with a bad metabolic and functional state background (depolarising excitation). There are also two types of inhibition: inhibition with a good metabolic and functional state background (hyperpolarising inhibition) and inhibition with a bad metabolic and functional state background (depolarising inhibition). The proper understanding of the processes in the nervous system makes it possible to approach the decision of a whole number of fundamental problems of psychophysiology.
doi:10.1016/j.ijpsycho.2010.06.216
On the functional states of nerve cells (a new approach) S.E. Murik Irkutsk State University, Russia What is the difference in the reaction of neurons to favorable and unfavorable irritants? It should be noted that current research has not yet yielded any results. Most research tends to regard excitation and inhibition as the two functional states of neurons. We find this claim to be false, as there is a substitution of one notion for another. From our point of view, excitation and inhibition are the working (operational) states, but not the functional states. We suggest that the term “functional state” should be employed only in those cases in which one has to characterize the current ability of the system to perform working (operating) actions. This ability may be described with qualitative adjectives such as “good” or “bad”, “better” or “worse”, “the best” (“excellent”) or “the worst”. Some functional states have specific names; for example, fatigue is a functional state. Besides, it is not clear how to distinguish a good functional state of a nerve cell from a bad functional state of one. At present, it seems impossible to differentiate excitation with a good functional state background from the one with a bad functional state background. The analysis of numerous scientific data showed that, at present, there are many reasons to consider the response of nerve cells to any irritant (i.e. unfavorable factors) as an adoptive reaction that has several stages. Alongside these changes in adoptive stages, changes in the degree of cell membrane polarization (i.e. membrane potential) occur. We distinguish four adaptive stages in the reaction of nerve cells to irritants and unfavorable factors (see Fig. 1): (I) hyperpolarisation (or hyperpolarising inhibition); (II) hyperpolarising (posthyperpolarising) excitation; (III) depolarising excitation; (IV) depolarising inhibition. Each of these stages is characterized by its own peculiarities of the elapse of living processes, the functional capacities of neurons, and also different resistance to unfavorable factors. In fact, there are two types of neuronal excitation: excitation with a good metabolic and functional state background (posthyperpolarising
Fig. 1. Variation in membrane potential (MP) and change in the state of neural tissue under the action of stimuli with time (t). The functional state goes through the following stages of change: 0 — rest, I — hyperpolarisation silence (“inhibition”), II — posthyperpolarisation excitation, III — depolarisation excitation, IV — depolarisation suppression (“inhibition”). An arrow indicates the beginning of the action of the stress-stimulus. RP — resting potential level.
doi:10.1016/j.ijpsycho.2010.06.217
High-performance computing system for electroencephalographic data analysis Zsolt I. Lázára, Jouke R. Heringab, István Pappa, Alpár S. Lázárc,d a Faculty of Physics, Babeş-Bolyai University, Cluj-Napoca, Romania b Faculty of Applied Sciences, Delft University of Technology, The Netherlands c Institute of Behavioural Sciences, Semmelweis University, Budapest, Hungary d Surrey Sleep Research Centre, University of Surrey, Guildford, UK Introduction: The emergence of novel methods in the analysis of time series and that of complex systems, e.g. networks, has opened new avenues into the interpretation of EEG signals. Most of these methods depend superlinearly on the size of the data. With sampling frequencies up to 1024 Hz, channels up to 128, recording times up to ten hours, and subject groups over 100, the size of primary data can reach terabytes. The optimal processing of this amount of data is a challenge both from the point of computational cycles and the I/O operations. Fortunately, the other factor contributing to the renaissance of EEG is the development of computer hardware technologies following Moore's exponential law unabatedly. However, a software system that can make use of today's high performance hardware infrastructures, such as computing clusters, while providing an “EEG friendly” user interface and involving only free and/or OpenSource software, is not available. Aims: Our first goal was to identify the specifics of EEG data analysis in comparison with other fields of scientific computing, such as molecular dynamics' simulations or analysis of experimental highenergy physics data. These specifics stand as the basis of the requirements of a software system, HIgh PERformance BiOsignaL Computing (HIPERBOL), that would optimally suit electroencephalo-
312
Abstracts / International Journal of Psychophysiology 77 (2010) 288–342
graphy requirements such as: reuse of existing free and/or OpenSource software running on commodity hardware, graphical job monitoring, transparency, job resubmission upon failure, redundancy and proper bookkeeping of data and algorithms, adaptability to different parallel computing environments (multicore PCs, clusters, computational grids), etc. The system should be designed following the above guidelines and implemented accordingly. Eventually, the system will have to be validated by applying it to real life projects. Results: Following the list of requirements, we created a software infrastructure on one of the computational clusters of the Physics Department at the Babes-Bolyai University in Cluj, Romania (84 cores on 11 nodes). Most of the code was reused from third party OpenSource projects, such as Scientific Linux, Oscar, Ganga, or Python, and supplemented with newly written modules for covering the missing features and integrating the different parts. The system has been successfully tested while assisting two projects carried out at two different laboratories of the Semmelweis University in Budapest, Hungary. A whole range of different quantitative, both linear and nonlinear, and statistical analysis algorithms have been employed on over twenty gigabytes of primary data. The efficiency of the system is best illustrated by the spectral and coherence analysis of a ten channel (45 channel pairs), whole night sleep set of EEG data from 31 subjects that required a little over a minute to complete. Conclusions: HIPERBOL proved to be feasible, useful, and highly efficient. Most of the code can be reused for processing fMRI data without major changes. In that case, the amount of data exceeds that of typical EEG recordings, and therefore, the relevance of the system would be even more pronounced. The user interfaces and the deployment of the system need further improvements so that it can be easily handled by users who are not computer experts. Given that the system is purely OpenSource, the contribution to the scientific community is both possible and desirable. doi:10.1016/j.ijpsycho.2010.06.218
Simultaneous recording of EEG and transcranial electric stimulation Toralf Neulinga, Tino Zaehleb, Christoph Herrmanna Experimental Psychology Lab, Carl von Ossietzky University, Oldenburg, Germany b Clinic for Neurology, Otto von Guericke Universtity, Magdeburg, Germany
a
Transcranial electric stimulation (TES) offers the possibility to non-invasively stimulate the human brain in various ways. For example, transcranial direct current stimulation (tDCS) has been demonstrated to both excite and inhibit neural tissue (Antal et al., 2006). However, it would be desirable to directly see the consequence of tDCS in simultaneously recorded electroencephalography (EEG). Therefore, we conducted a series of three experiments, all simultaneously recording EEG during TES. The dynamic range of the EEG-amplifier was chosen in order to fully sample the amplitude of the stimulation artifact. This offers the possibility to subtract the artifact from the EEG and subsequently, to analyze the EEG-response to TES. In the first experiment 10 min tDCS was applied in combination with a visual working memory task. The results demontrate that anodal tDCS over the left dorsolateral prefrontal cortex enhances EEG alpha- and theta activities and at the same time improves memory performance. In the second experiment, we applied transcranial alternating current stimulation (tACS), which was shown to have consequences on perception (Kanai et al., 2008) and memory performance (Marshall et al., 2006). We applied tACS at individual EEG alpha frequency for 10 min. EEG spectra recorded before and
after tACS show enhanced alpha activity only for the stimulation group but not for the sham group. In a third experiment, we applied single pulses of tDCS revealing enhanced alpha activity, probably resulting from a phase reset of prestimulus alpha oscillations. All three experiments demonstrate the feasibility of simultaneously recording EEG during transcranial electrical stimulation (TES) which will probably become a valuable tool in neuroscience in the near future. References Antal, A., Nitsche, M.A., Paulus, W., 2006. Transcranial direct current stimulation and the visual cortexBrain Res Bull 68, 459–463. Kanai, R., Chaieb, L., Antal, A., Walsh, V., Paulus, W., 2008. Frequency-dependent electrical stimulation of the visual cortex. Curr Biol 18, 1839–1843. Marshall, L., Helgadottir, H., Molle, M., Born, J., 2006. Boosting slow oscillations during sleep potentiates memory. Nature 444, 610–613.
doi:10.1016/j.ijpsycho.2010.06.219
Human oscillatory activities associated with reward and punishment processing in a probabilistic reversal learning task Bai Yua,b, Hideki Ohiraa a Department of Psychology, Graduate School of Environmental Studies, Nagoya University, Nagoya, Japan b Japan Society for the Promotion of Science, Tokyo, Japan Optimal behavior in a competitive world requires the flexibility to adapt decision strategies based on recent outcomes. The Feedback Error-Related Negativity (FRN) is an event-related potential (ERP) modulation that distinguishes losses from wins. Following the FRN, another ERP component is usually observed (error-positivity: Pe), which reflects awareness of the error, but little is known about the effects of outcome probability on these ERP responses. In the present study, we investigated the effects of outcome probability on the FRN and the frequency characteristics of feedback processing by a probabilistic reversal learning task. Participants were 15 healthy graduate and undergraduate students (M = 20.1 SD = 1.4). Written informed consent was obtained from all participants. We recorded event-related brain potentials (ERPs) while subjects performed the probabilistic reversal learning task. In the task, subjects saw two abstract line drawings on the left and right sides of the screen, and had to choose one on each trial, after which they either won or lost 30 Yen (JPY). In 80 trials of a block, one stimulus was associated with reward at a probability of 70% but with punishment at a probability of 30% (advantageous stimulus), and the other stimulus was associated with reward at a probability of 30% and with punishment at a probability of 70% (disadvantageous stimulus). After 80 trials, the contingencies were reversed. The first 40 trials of the first half of one block were defined as ‘learning period’ and the last 40 trials were defined as ‘adapted period’. The first 40 trials of the latter half of one block were defined as ‘reversal learning period’ and the last 40 trials were defined as ‘reversal adapted period’. Behaviorally, all subjects quickly adapted their decision-making to maximize rewards. The amplitude of Pe for the ‘advantageous stimulus’ was significantly larger in the ‘learning period’, but that of FRN was not. This would suggest that the Pe was influenced by the outcome probability. These results support the assumption that Pe may reflect an evaluation for outcomes of decision-making on a relatively longer time scale. Compared to Pe, the FRN may reflect an evaluation of outcomes on a shorter time scale. Consistent with the previous studies, in the present experiment, we also observed enhanced theta power in the loss condition and enhanced beta power in the gain condition. These results showed that the reward-
stuttering is not an insulated (only peripheral) speech disorder, but it is part of a syndrome of particular mnestic, neurodynamic and motor defects revealing a dysfunction of deep posterior frontal and middle structures of the brain (functional blocks I and III according to Luria), and comprehensive neuropsychological and psychotherapeutic treatment can reduce these problems. 2. Psychophysiological results. Discrepancy in the potential of speech readiness in normally speaking and stuttering subjects is related to the pathological overactivation of the limbic system and frontal cortex and pathological deactivation of the temporal cortex, brainstem and basal nuclei. Also, the coordination of hemispheres in the process of preparation for speech is very weak in stutterers. Treatment of stuttering changes frontal activity, reduces limbic activity and reactivates the brainstem and temporal cortex. Generally, the longitudinal study revealed the changes of many psychophysiological parameters after speech rehabilitation including reintegration of activated and deactivated areas into a more normal pattern of brain activity and improvement of inter- and intrahemispheric interaction. Conclusion: Collected data allow us to update our knowledge about general neurocognitive mechanisms of stuttering and, particularly, about mechanisms of readiness for speech in a specific kind of distributed functional verbal system, such as stuttering.
311
excitation) and excitation with a bad metabolic and functional state background (depolarising excitation). There are also two types of inhibition: inhibition with a good metabolic and functional state background (hyperpolarising inhibition) and inhibition with a bad metabolic and functional state background (depolarising inhibition). The proper understanding of the processes in the nervous system makes it possible to approach the decision of a whole number of fundamental problems of psychophysiology.
doi:10.1016/j.ijpsycho.2010.06.216
On the functional states of nerve cells (a new approach) S.E. Murik Irkutsk State University, Russia What is the difference in the reaction of neurons to favorable and unfavorable irritants? It should be noted that current research has not yet yielded any results. Most research tends to regard excitation and inhibition as the two functional states of neurons. We find this claim to be false, as there is a substitution of one notion for another. From our point of view, excitation and inhibition are the working (operational) states, but not the functional states. We suggest that the term “functional state” should be employed only in those cases in which one has to characterize the current ability of the system to perform working (operating) actions. This ability may be described with qualitative adjectives such as “good” or “bad”, “better” or “worse”, “the best” (“excellent”) or “the worst”. Some functional states have specific names; for example, fatigue is a functional state. Besides, it is not clear how to distinguish a good functional state of a nerve cell from a bad functional state of one. At present, it seems impossible to differentiate excitation with a good functional state background from the one with a bad functional state background. The analysis of numerous scientific data showed that, at present, there are many reasons to consider the response of nerve cells to any irritant (i.e. unfavorable factors) as an adoptive reaction that has several stages. Alongside these changes in adoptive stages, changes in the degree of cell membrane polarization (i.e. membrane potential) occur. We distinguish four adaptive stages in the reaction of nerve cells to irritants and unfavorable factors (see Fig. 1): (I) hyperpolarisation (or hyperpolarising inhibition); (II) hyperpolarising (posthyperpolarising) excitation; (III) depolarising excitation; (IV) depolarising inhibition. Each of these stages is characterized by its own peculiarities of the elapse of living processes, the functional capacities of neurons, and also different resistance to unfavorable factors. In fact, there are two types of neuronal excitation: excitation with a good metabolic and functional state background (posthyperpolarising
Fig. 1. Variation in membrane potential (MP) and change in the state of neural tissue under the action of stimuli with time (t). The functional state goes through the following stages of change: 0 — rest, I — hyperpolarisation silence (“inhibition”), II — posthyperpolarisation excitation, III — depolarisation excitation, IV — depolarisation suppression (“inhibition”). An arrow indicates the beginning of the action of the stress-stimulus. RP — resting potential level.
doi:10.1016/j.ijpsycho.2010.06.217
High-performance computing system for electroencephalographic data analysis Zsolt I. Lázára, Jouke R. Heringab, István Pappa, Alpár S. Lázárc,d a Faculty of Physics, Babeş-Bolyai University, Cluj-Napoca, Romania b Faculty of Applied Sciences, Delft University of Technology, The Netherlands c Institute of Behavioural Sciences, Semmelweis University, Budapest, Hungary d Surrey Sleep Research Centre, University of Surrey, Guildford, UK Introduction: The emergence of novel methods in the analysis of time series and that of complex systems, e.g. networks, has opened new avenues into the interpretation of EEG signals. Most of these methods depend superlinearly on the size of the data. With sampling frequencies up to 1024 Hz, channels up to 128, recording times up to ten hours, and subject groups over 100, the size of primary data can reach terabytes. The optimal processing of this amount of data is a challenge both from the point of computational cycles and the I/O operations. Fortunately, the other factor contributing to the renaissance of EEG is the development of computer hardware technologies following Moore's exponential law unabatedly. However, a software system that can make use of today's high performance hardware infrastructures, such as computing clusters, while providing an “EEG friendly” user interface and involving only free and/or OpenSource software, is not available. Aims: Our first goal was to identify the specifics of EEG data analysis in comparison with other fields of scientific computing, such as molecular dynamics' simulations or analysis of experimental highenergy physics data. These specifics stand as the basis of the requirements of a software system, HIgh PERformance BiOsignaL Computing (HIPERBOL), that would optimally suit electroencephalo-
312
Abstracts / International Journal of Psychophysiology 77 (2010) 288–342
graphy requirements such as: reuse of existing free and/or OpenSource software running on commodity hardware, graphical job monitoring, transparency, job resubmission upon failure, redundancy and proper bookkeeping of data and algorithms, adaptability to different parallel computing environments (multicore PCs, clusters, computational grids), etc. The system should be designed following the above guidelines and implemented accordingly. Eventually, the system will have to be validated by applying it to real life projects. Results: Following the list of requirements, we created a software infrastructure on one of the computational clusters of the Physics Department at the Babes-Bolyai University in Cluj, Romania (84 cores on 11 nodes). Most of the code was reused from third party OpenSource projects, such as Scientific Linux, Oscar, Ganga, or Python, and supplemented with newly written modules for covering the missing features and integrating the different parts. The system has been successfully tested while assisting two projects carried out at two different laboratories of the Semmelweis University in Budapest, Hungary. A whole range of different quantitative, both linear and nonlinear, and statistical analysis algorithms have been employed on over twenty gigabytes of primary data. The efficiency of the system is best illustrated by the spectral and coherence analysis of a ten channel (45 channel pairs), whole night sleep set of EEG data from 31 subjects that required a little over a minute to complete. Conclusions: HIPERBOL proved to be feasible, useful, and highly efficient. Most of the code can be reused for processing fMRI data without major changes. In that case, the amount of data exceeds that of typical EEG recordings, and therefore, the relevance of the system would be even more pronounced. The user interfaces and the deployment of the system need further improvements so that it can be easily handled by users who are not computer experts. Given that the system is purely OpenSource, the contribution to the scientific community is both possible and desirable. doi:10.1016/j.ijpsycho.2010.06.218
Simultaneous recording of EEG and transcranial electric stimulation Toralf Neulinga, Tino Zaehleb, Christoph Herrmanna Experimental Psychology Lab, Carl von Ossietzky University, Oldenburg, Germany b Clinic for Neurology, Otto von Guericke Universtity, Magdeburg, Germany
a
Transcranial electric stimulation (TES) offers the possibility to non-invasively stimulate the human brain in various ways. For example, transcranial direct current stimulation (tDCS) has been demonstrated to both excite and inhibit neural tissue (Antal et al., 2006). However, it would be desirable to directly see the consequence of tDCS in simultaneously recorded electroencephalography (EEG). Therefore, we conducted a series of three experiments, all simultaneously recording EEG during TES. The dynamic range of the EEG-amplifier was chosen in order to fully sample the amplitude of the stimulation artifact. This offers the possibility to subtract the artifact from the EEG and subsequently, to analyze the EEG-response to TES. In the first experiment 10 min tDCS was applied in combination with a visual working memory task. The results demontrate that anodal tDCS over the left dorsolateral prefrontal cortex enhances EEG alpha- and theta activities and at the same time improves memory performance. In the second experiment, we applied transcranial alternating current stimulation (tACS), which was shown to have consequences on perception (Kanai et al., 2008) and memory performance (Marshall et al., 2006). We applied tACS at individual EEG alpha frequency for 10 min. EEG spectra recorded before and
after tACS show enhanced alpha activity only for the stimulation group but not for the sham group. In a third experiment, we applied single pulses of tDCS revealing enhanced alpha activity, probably resulting from a phase reset of prestimulus alpha oscillations. All three experiments demonstrate the feasibility of simultaneously recording EEG during transcranial electrical stimulation (TES) which will probably become a valuable tool in neuroscience in the near future. References Antal, A., Nitsche, M.A., Paulus, W., 2006. Transcranial direct current stimulation and the visual cortexBrain Res Bull 68, 459–463. Kanai, R., Chaieb, L., Antal, A., Walsh, V., Paulus, W., 2008. Frequency-dependent electrical stimulation of the visual cortex. Curr Biol 18, 1839–1843. Marshall, L., Helgadottir, H., Molle, M., Born, J., 2006. Boosting slow oscillations during sleep potentiates memory. Nature 444, 610–613.
doi:10.1016/j.ijpsycho.2010.06.219
Human oscillatory activities associated with reward and punishment processing in a probabilistic reversal learning task Bai Yua,b, Hideki Ohiraa a Department of Psychology, Graduate School of Environmental Studies, Nagoya University, Nagoya, Japan b Japan Society for the Promotion of Science, Tokyo, Japan Optimal behavior in a competitive world requires the flexibility to adapt decision strategies based on recent outcomes. The Feedback Error-Related Negativity (FRN) is an event-related potential (ERP) modulation that distinguishes losses from wins. Following the FRN, another ERP component is usually observed (error-positivity: Pe), which reflects awareness of the error, but little is known about the effects of outcome probability on these ERP responses. In the present study, we investigated the effects of outcome probability on the FRN and the frequency characteristics of feedback processing by a probabilistic reversal learning task. Participants were 15 healthy graduate and undergraduate students (M = 20.1 SD = 1.4). Written informed consent was obtained from all participants. We recorded event-related brain potentials (ERPs) while subjects performed the probabilistic reversal learning task. In the task, subjects saw two abstract line drawings on the left and right sides of the screen, and had to choose one on each trial, after which they either won or lost 30 Yen (JPY). In 80 trials of a block, one stimulus was associated with reward at a probability of 70% but with punishment at a probability of 30% (advantageous stimulus), and the other stimulus was associated with reward at a probability of 30% and with punishment at a probability of 70% (disadvantageous stimulus). After 80 trials, the contingencies were reversed. The first 40 trials of the first half of one block were defined as ‘learning period’ and the last 40 trials were defined as ‘adapted period’. The first 40 trials of the latter half of one block were defined as ‘reversal learning period’ and the last 40 trials were defined as ‘reversal adapted period’. Behaviorally, all subjects quickly adapted their decision-making to maximize rewards. The amplitude of Pe for the ‘advantageous stimulus’ was significantly larger in the ‘learning period’, but that of FRN was not. This would suggest that the Pe was influenced by the outcome probability. These results support the assumption that Pe may reflect an evaluation for outcomes of decision-making on a relatively longer time scale. Compared to Pe, the FRN may reflect an evaluation of outcomes on a shorter time scale. Consistent with the previous studies, in the present experiment, we also observed enhanced theta power in the loss condition and enhanced beta power in the gain condition. These results showed that the reward-