Multi-dimensional time series classification for early epilepsy diagnosis 12:00-01:30 Lunch Break

Epilepsy is one of the most common chronic neurological disorders that afflicts close to 50 million people worldwide. Antiepilepsy drugs (AEDs), the mainstay of epilepsy treatment, control seizures in two thirds of patients only. Other therapies include the ketogenic diet, ablative surgery, hormonal treatments and neurostimulation. While other approaches to stimulation of the brain are currently in the experimental phase vagus nerve stimulation (VNS) has been approved by the FDA since July 1997 for the adjunctive treatment of intractable partial onset epilepsy with and without secondary generalization in patients twelve years of age or older. The safety and efficacy of VNS have been proven and duplicated in two subsequent double-blinded controlled studies after two pilot studies demonstrated the feasibility of VNS in man. Long term observational studies confirmed the safety of VNS and that its effectiveness is sustained over time. While AEDs influence seizure thresholds via blockade or modulation of ionic channels, inhibit excitatory neurotransmitters or enhance inhibitory neurotransmitters the exact mechanism of action of VNS is not known. Neuroimaging studies revealed that VNS increases blood flow in certain regions of the brain such as the thalamus. Chemical lesions in the rats brain showed that norepinephrine is an important link in the anticonvulsant effect of VNS. Analysis of cerebrospinal fluid obtained from patients before and after treatment with VNS showed modest decreases in excitatory neurotransmitters. Although Hammond et al reported no effect of VNS on scalp EEG by visual analysis and Salinsky et al found no effect of VNS on scalp EEG by spectral analysis, Kuba et al suggested that VNS reduces interictal epileptiform activity. Further, nonlinear dynamical analysis of the electroencephalogram in the rat and man have reportedly shown predictable changes (decrease in the short term Lyapunov exponent STLmax and T-index) more than an hour prior to the clinical or electroencephalographic seizure onset (Iasemidis et al, 2003). It is possible that intermittent VNS maintains chaoticity of brain activity in patients with epilepsy that respond to this therapy. The most optimal stimulation parameters of VNS are not known and further study of nonlinear dynamics of brain activity may shed some light on more effective interception or prevention of seizures. Online real time analysis may allow on-demand stimulation rather than hit-or-miss approach. Additionally, it may assist clinicians in identifying seizure ictal events in challenging situations where the borderland between nonconvulsive status epilepticus and other conditions of encephalopathy using the best diagnostic test, the electroencephalogram, is, at best, blurry. ? email: [email protected] Support vector machine classification of EEG data from epileptic patients treated with vagus nerve stimulation Michael A. Bewernitz, George Ghacibeh, Onur Seref, Basim Uthman, and Panos M. Pardalos 1 Department of Biomedical Engineering University of Florida Gainesville, FL 32611 2 Department of Neurology McKnight Brain Institute University of Florida Gainesville, FL 32610 Abstract. In this presentation we describe an application of support vector machines (SVMs) to the analysis of electroencephalogram (EEG) obtained from the scalp of epileptic patients implanted with the vagus nerve stimulator (VNS). The purpose of this study is to devise a physiologic marker using scalp EEG for optimal VNS parameters. Scalp EEG recordings were obtained from four epileptic patients with VNS implants. Averaged scalp EEG samples were used as features for separation. SVM classification was used to separate a time segment during the beginning of stimulation from all the successive non-overlapping time segments within a full VNS on/off cycle. This analysis was performed for all the automated VNS cycles occurring during approximately twenty-four hours of scalp EEG. These results suggest that successful VNS therapy provides seizure protection by rapidly altering the state of the brain following stimulation. These findings also imply that VNS stimulation is less likely to exert full therapeutic effect if stimulations do not consistently produce significant changes in the EEG. These findings could provide a foundation for an electrographic marker of optimal VNS parameters. In this presentation we describe an application of support vector machines (SVMs) to the analysis of electroencephalogram (EEG) obtained from the scalp of epileptic patients implanted with the vagus nerve stimulator (VNS). The purpose of this study is to devise a physiologic marker using scalp EEG for optimal VNS parameters. Scalp EEG recordings were obtained from four epileptic patients with VNS implants. Averaged scalp EEG samples were used as features for separation. SVM classification was used to separate a time segment during the beginning of stimulation from all the successive non-overlapping time segments within a full VNS on/off cycle. This analysis was performed for all the automated VNS cycles occurring during approximately twenty-four hours of scalp EEG. These results suggest that successful VNS therapy provides seizure protection by rapidly altering the state of the brain following stimulation. These findings also imply that VNS stimulation is less likely to exert full therapeutic effect if stimulations do not consistently produce significant changes in the EEG. These findings could provide a foundation for an electrographic marker of optimal VNS parameters. ? email: [email protected] Biclustering EEG Data from Epileptic Patients Treated with Vagus Nerve Stimulation Nikita Boyko , George Ghacibeh, Michael Bewernitz, Stanislav Busygin, and Panos M. Pardalos 1 Department of Industrial and Systems Engineering University of Florida Gainesville, FL 32611 2 Department of Biomedical Engineering University of Florida Gainesville, FL 32611 3 Department of Neurology McKnight Brain Institute University of Florida Gainesville, FL 32610 Abstract. We present a pilot study of an application of consistent biclustering to analyze scalp EEG data obtained from epileptic patients undergoing treatment with a vagus nerve stimulator (VNS). The ultimate goal of this study is to develop a physiologic marker for optimal VNS parameters (e.g. output current, signal frequency, etc.) using measures of scalp EEG signals. A time series of STLmax values was computed for each scalp EEG channel recorded from two epileptic patients and used as a feature of the two datasets. The averaged samples from stimulation periods were then separated from averaged samples from non-stimulation periods by feature selection performed within the consistent biclustering routine. The obtained biclustering results allow us to assume that signals from certain parts of the brain consistently change their characteristics when VNS is switched on and could provide a basis for desirable VNS stimulation parameters. A physiologic marker of optimal VNS effect could greatly reduce the cost, time, and risk of calibrating VNS stimulation parameters in newly implanted patients compared to the current method of clinical response. We present a pilot study of an application of consistent biclustering to analyze scalp EEG data obtained from epileptic patients undergoing treatment with a vagus nerve stimulator (VNS). The ultimate goal of this study is to develop a physiologic marker for optimal VNS parameters (e.g. output current, signal frequency, etc.) using measures of scalp EEG signals. A time series of STLmax values was computed for each scalp EEG channel recorded from two epileptic patients and used as a feature of the two datasets. The averaged samples from stimulation periods were then separated from averaged samples from non-stimulation periods by feature selection performed within the consistent biclustering routine. The obtained biclustering results allow us to assume that signals from certain parts of the brain consistently change their characteristics when VNS is switched on and could provide a basis for desirable VNS stimulation parameters. A physiologic marker of optimal VNS effect could greatly reduce the cost, time, and risk of calibrating VNS stimulation parameters in newly implanted patients compared to the current method of clinical response. ? email: [email protected] Engineering synthetic killer circuits in bacteria Lingchong You ? Department of Biomedical Engineering Institute for Genome Sciences & Policy Duke University Abstract. Synthetic biology, which often entails de novo engineering of gene circuits with well-defined function, has shown great potential to impact diverse areas. One appealing application is the reprogramming of bacteria as therapeutic agents, for example, to deliver drugs or to selectively kill tumor cells. To realize this goal, however, we must be able to precisely control bacterial dynamics, including growth, death, and aggregation, under diverse conditions. Combining modeling and experiment, my laboratory has been exploring design strategies to achieve such control by engineering a series of synthetic killer circuits in bacterium Escherichia coli. Here I will discuss our recent progress in this direction. Based on our experience, I will comment on major challenges involved in the engineering process, and how such design exercise can also provide valuable insights into behavior of natural biological systems. Synthetic biology, which often entails de novo engineering of gene circuits with well-defined function, has shown great potential to impact diverse areas. One appealing application is the reprogramming of bacteria as therapeutic agents, for example, to deliver drugs or to selectively kill tumor cells. To realize this goal, however, we must be able to precisely control bacterial dynamics, including growth, death, and aggregation, under diverse conditions. Combining modeling and experiment, my laboratory has been exploring design strategies to achieve such control by engineering a series of synthetic killer circuits in bacterium Escherichia coli. Here I will discuss our recent progress in this direction. Based on our experience, I will comment on major challenges involved in the engineering process, and how such design exercise can also provide valuable insights into behavior of natural biological systems. ? email: [email protected] Multi-dimensional time series classification for early epilepsy diagnosis Wanpracha Art Chaovalitwongse ? Industrial and Systems Engineering Rutgers University Piscataway, NJ 08854 Abstract. Typically neurologists study the brain function through neurophysiological signals like Electroencephalograms (EEGs), in which spatial and temporal properties are encrypted. Analyzing these massive data from the brain to identify the abnormalities or anomalies is extremely challenging. In this study, we propose a novel multi-dimensional time series classication technique, namely connectivity support vector machine (C-SVM), to identify abnormal brain patterns specic to epilepsy disease based on scalp EEG signals acquired from normal and epilepsy patients. C-SVM is an integration of brain-connectivity network modeling and support vector machine (SVM). The ultimate goal of this study is to develop and validate a quick and accurate epilepsy screening framework for initial brain diagnosis. The data set used in this study is multi-channel EEG scalp recordings acquired from 5 normal and 5 epilepsy patients. Generally, scalp EEG data are contaminated by noises (e.g., movement, artifacts). Two independent component analysis (ICA) algorithms, Thin Algorithm and Un-biased Quasi Newton Method, are applied to filter noises from the EEG data. The empirical results indicated that, performing 5fold cross validation, C-SVM achieved 94.6identifying epilepsy patients compared to 78standard SVM. Typically neurologists study the brain function through neurophysiological signals like Electroencephalograms (EEGs), in which spatial and temporal properties are encrypted. Analyzing these massive data from the brain to identify the abnormalities or anomalies is extremely challenging. In this study, we propose a novel multi-dimensional time series classication technique, namely connectivity support vector machine (C-SVM), to identify abnormal brain patterns specic to epilepsy disease based on scalp EEG signals acquired from normal and epilepsy patients. C-SVM is an integration of brain-connectivity network modeling and support vector machine (SVM). The ultimate goal of this study is to develop and validate a quick and accurate epilepsy screening framework for initial brain diagnosis. The data set used in this study is multi-channel EEG scalp recordings acquired from 5 normal and 5 epilepsy patients. Generally, scalp EEG data are contaminated by noises (e.g., movement, artifacts). Two independent component analysis (ICA) algorithms, Thin Algorithm and Un-biased Quasi Newton Method, are applied to filter noises from the EEG data. The empirical results indicated that, performing 5fold cross validation, C-SVM achieved 94.6identifying epilepsy patients compared to 78standard SVM. ? email: [email protected] Quantitative analysis for electrooculography (EOG) for neurodegenerative diseases Petros Xanthopoulos , Chang-Chia Liu, Onur Seref, Alla Kammerdiner, Panos Pardalos, Chris Sackellares, and Frank Skidmore 1 Department of Industrial and Systems Engineering University of Florida Gainesville, FL 32611 2 Department of Biomedical Engineering University of Florida Gainesville, FL 32611 3 Department of Neurology McKnight Brain Institute University of Florida Gainesville, FL 32610 Abstract. Many studies have documented the irregular motions for horizontal and vertical saccades in several human neurodegenerative diseases. Eye movement measurement plays an important role in the examination the progress of the neurodegenerative diseases. There are several techniques for measuring eye movement, the Infrared detection technique (IR), Video-oculography (VOG), Scleral eye coil and EOG. Among those available detection techniques, EOG is a major source for monitoring the abnormal eye movement. In this real-time quantitative analysis study, the methods which can capture the characteristic of the eye movement were proposed to accurately categorize the state of neurodegenerative subjects. The EOG recordings were taken while tested subjects were watching a short (¿120s) animation clip. The animation clip was designed to execute the eyeballs of the subject in different motions such as vertical smooth pursued (SVP), horizontal smooth pursued (HVP) and random saccades (RS). Five normal volunteer subjects were included in the study; the real time quantitative analysis was implemented using proposed methods. The statistics of the EOG recording from the five normal subjects will serve to conduct the test statistic. Diagnosis of abnormal oculus motions is an important step in evaluation of patients with developmental neurodegenerative diseases. A standard real-time quantitative analysis will not only improve the accuracy of doctors examination and also provide a better understanding of various pathologies. Many studies have documented the irregular motions for horizontal and vertical saccades in several human neurodegenerative diseases. Eye movement measurement plays an important role in the examination the progress of the neurodegenerative diseases. There are several techniques for measuring eye movement, the Infrared detection technique (IR), Video-oculography (VOG), Scleral eye coil and EOG. Among those available detection techniques, EOG is a major source for monitoring the abnormal eye movement. In this real-time quantitative analysis study, the methods which can capture the characteristic of the eye movement were proposed to accurately categorize the state of neurodegenerative subjects. The EOG recordings were taken while tested subjects were watching a short (¿120s) animation clip. The animation clip was designed to execute the eyeballs of the subject in different motions such as vertical smooth pursued (SVP), horizontal smooth pursued (HVP) and random saccades (RS). Five normal volunteer subjects were included in the study; the real time quantitative analysis was implemented using proposed methods. The statistics of the EOG recording from the five normal subjects will serve to conduct the test statistic. Diagnosis of abnormal oculus motions is an important step in evaluation of patients with developmental neurodegenerative diseases. A standard real-time quantitative analysis will not only improve the accuracy of doctors examination and also provide a better understanding of various pathologies.