@article {3405, title = {Localizing ECoG electrodes on the cortical anatomy without post-implantation imaging.}, journal = {Neuroimage Clin}, volume = {6}, year = {2014}, month = {08/2014}, pages = {64-76}, abstract = {

INTRODUCTION: Electrocorticographic (ECoG) grids are placed subdurally on the cortex in people undergoing cortical resection to delineate eloquent cortex. ECoG signals have high spatial and temporal resolution and thus can be valuable for neuroscientific research. The value of these data is highest when they can be related to the cortical anatomy. Existing methods that establish this relationship rely either on post-implantation imaging using computed tomography (CT), magnetic resonance imaging (MRI) or X-Rays, or on intra-operative photographs. For research purposes, it is desirable to localize ECoG electrodes on the brain anatomy even when post-operative imaging is not available or when intra-operative photographs do not readily identify anatomical landmarks.

METHODS: We developed a method to co-register ECoG electrodes to the underlying cortical anatomy using only a pre-operative MRI, a clinical neuronavigation device (such as BrainLab VectorVision), and fiducial markers. To validate our technique, we compared our results to data collected from six subjects who also had post-grid implantation imaging available. We compared the electrode coordinates obtained by our fiducial-based method to those obtained using existing methods, which are based on co-registering pre- and post-grid implantation images.

RESULTS: Our fiducial-based method agreed with the MRI-CT method to within an average of 8.24 mm (mean, median = 7.10 mm) across 6 subjects in 3 dimensions. It showed an average discrepancy of 2.7 mm when compared to the results of the intra-operative photograph method in a 2D coordinate system. As this method does not require post-operative imaging such as CTs, our technique should prove useful for research in intra-operative single-stage surgery scenarios. To demonstrate the use of our method, we applied our method during real-time mapping of eloquent cortex during a single-stage surgery. The results demonstrated that our method can be applied intra-operatively in the absence of post-operative imaging to acquire ECoG signals that can be valuable for neuroscientific investigations.

}, keywords = {auditory processing, electrocorticography (ECoG), electrode localization, fiducials, interaoperative localization}, issn = {2213-1582}, doi = {10.1016/j.nicl.2014.07.015}, url = {http://www.ncbi.nlm.nih.gov/pubmed/25379417}, author = {Disha Gupta and Jeremy Jeremy Hill and Adamo, Matthew A and A L Ritaccio and Gerwin Schalk} } @article {3361, title = {Simultaneous Real-Time Monitoring of Multiple Cortical Systems.}, journal = {Journal of Neural Engineering}, year = {2014}, month = {10/2014}, abstract = {OBJECTIVE: Real-time monitoring of the brain is potentially valuable for performance monitoring, communication, training or rehabilitation. In natural situations, the brain performs a complex mix of various sensory, motor or cognitive functions. Thus, real-time brain monitoring would be most valuable if (a) it could decode information from multiple brain systems simultaneously, and (b) this decoding of each brain system were robust to variations in the activity of other (unrelated) brain systems. Previous studies showed that it is possible to decode some information from different brain systems in retrospect and/or in isolation. In our study, we set out to determine whether it is possible to simultaneously decode important information about a user from different brain systems in real time, and to evaluate the impact of concurrent activity in different brain systems on decoding performance. APPROACH: We study these questions using electrocorticographic signals recorded in humans. We first document procedures for generating stable decoding models given little training data, and then report their use for offline and for real-time decoding from 12 subjects (six for offline parameter optimization, six for online experimentation). The subjects engage in tasks that involve movement intention, movement execution and auditory functions, separately, and then simultaneously. Main Results: Our real-time results demonstrate that our system can identify intention and movement periods in single trials with an accuracy of 80.4\% and 86.8\%, respectively (where 50\% would be expected by chance). Simultaneously, the decoding of the power envelope of an auditory stimulus resulted in an average correlation coefficient of 0.37 between the actual and decoded power envelopes. These decoders were trained separately and executed simultaneously in real time. SIGNIFICANCE: This study yielded the first demonstration that it is possible to decode simultaneously the functional activity of multiple independent brain systems. Our comparison of univariate and multivariate decoding strategies, and our analysis of the influence of their decoding parameters, provides benchmarks and guidelines for future research on this topic.}, keywords = {auditory processing, Electrocorticography, movement intention, realtime decoding, simultaneous decoding}, doi = {10.1088/1741-2560/11/5/056001}, url = {http://www.ncbi.nlm.nih.gov/pubmed/25080161}, author = {Disha Gupta and Jeremy Jeremy Hill and Peter Brunner and Gunduz, Aysegul and A L Ritaccio and Gerwin Schalk} } @article {3377, title = {Novel inter-hemispheric white matter connectivity in the BTBR mouse model of autism.}, journal = {Brain Res}, volume = {1513}, year = {2013}, month = {06/2013}, pages = {26-33}, abstract = {Alterations in the volume, density, connectivity and functional activation of white matter tracts are reported in some individuals with autism and may contribute to their abnormal behaviors. The BTBR (BTBR T+tf/J) inbred strain of mouse, is used to model facets of autism because they develop low social behaviors, stereotypical and immune changes similar to those found in people with autism. Previously, it was thought a total absence of corpus callosal interhemispheric connective tissues in the BTBR mice may underlie their abnormal behaviors. However, postnatal lesions of the corpus callosum do not precipitate social behavioral problems in other strains of mice suggesting a flaw in this theory. In this study we used digital pathological methods to compare subcortical white matter connective tracts in the BTBR strain of mice with those found in the C57Bl/6 mouse and those reported in a standardized mouse brain atlas. We report, for the first time, a novel connective subcortical interhemispheric bridge of tissue in the posterior, but not anterior, cerebrum of the BTBR mouse. These novel connective tissues are comprised of myelinated fibers, with reduced myelin basic protein levels (MBP) compared to levels in the C57Bl/6 mouse. We used electrophysiological analysis and found increased inter-hemispheric connectivity in the posterior hemispheres of the BTBR strain compared with the anterior hemispheres. The conduction velocity was slower than that reported in normal mice. This study shows there is novel abnormal interhemispheric connectivity in the BTBR strain of mice, which may contribute to their behavioral abnormalities.}, keywords = {Analysis of Variance, Animals, Autistic Disorder, Brain, Corpus Callosum, Disease Models, Animal, Electroencephalography, Enzyme-Linked Immunosorbent Assay, Female, Functional Laterality, Image Processing, Computer-Assisted, Male, Mice, Mice, Inbred C57BL, Mice, Neurologic Mutants, Microtubule-Associated Proteins, Myelin Basic Protein, Nerve Fibers, Myelinated, Neuroimaging, Spectrum Analysis}, issn = {1872-6240}, doi = {10.1016/j.brainres.2013.04.001}, url = {http://www.ncbi.nlm.nih.gov/pubmed/23570707}, author = {Miller, V M and Disha Gupta and Neu, N and Cotroneo, A and Chadwick B. Boulay and Seegal, R F} } @article {2854, title = {ECoG: A Step Closer to the Brain @ The Brain Computer Interfacing Workshop, University Old dominion, Norfolk, VA}, year = {2012}, month = {05/2012}, author = {Disha Gupta} } @article {2924, title = {Proceedings of the Third International Workshop on Advances in Electrocorticography.}, journal = {Epilepsy Behav}, volume = {25}, year = {2012}, month = {12/2012}, pages = {605-13}, abstract = {The Third International Workshop on Advances in Electrocorticography (ECoG) was convened in Washington, DC, on November 10-11, 2011. As in prior meetings, a true multidisciplinary fusion of clinicians, scientists, and engineers from many disciplines gathered to summarize contemporary experiences in brain surface recordings. The proceedings of this meeting serve as evidence of a very robust and transformative field but will yet again require revision to incorporate the advances that the following year will surely bring.}, keywords = {Brain Mapping, brain-computer interface, Electrocorticography, Gamma-frequency electroencephalography, high-frequency oscillation, Neuroprosthetics, Seizure detection, Subdural grid}, issn = {1525-5069}, doi = {10.1016/j.yebeh.2012.09.016}, url = {http://www.ncbi.nlm.nih.gov/pubmed/23160096}, author = {A L Ritaccio and Beauchamp, Michael and Bosman, Conrado and Peter Brunner and Chang, Edward and Nathan E. Crone and Gunduz, Aysegul and Disha Gupta and Robert T. Knight and Leuthardt, Eric and Litt, Brian and Moran, Daniel and Ojemann, Jeffrey and Parvizi, Josef and Ramsey, Nick and Rieger, Jochem and Viventi, Jonathan and Voytek, Bradley and Williams, Justin and Gerwin Schalk} } @article {2097, title = {Recording Human Electrocorticographic (ECoG) Signals for Neuroscientific Research and Real-time Functional Cortical Mapping.}, journal = {J Vis Exp}, year = {2012}, month = {05/2012}, abstract = {

Neuroimaging studies of human cognitive, sensory, and motor processes are usually based on noninvasive techniques such as electroencephalography (EEG), magnetoencephalography or functional magnetic-resonance imaging. These techniques have either inherently low temporal or low spatial resolution, and suffer from low signal-to-noise ratio and/or poor high-frequency sensitivity. Thus, they are suboptimal for exploring the short-lived spatio-temporal dynamics of many of the underlying brain processes. In contrast, the invasive technique of electrocorticography (ECoG) provides brain signals that have an exceptionally high signal-to-noise ratio, less susceptibility to artifacts than EEG, and a high spatial and temporal resolution (i.e., \<1 cm/\<1 millisecond, respectively). ECoG involves measurement of electrical brain signals using electrodes that are implanted subdurally on the surface of the brain. Recent studies have shown that ECoG amplitudes in certain frequency bands carry substantial information about task-related activity, such as motor execution and planning,\ auditory\ processing and visual-spatial attention. Most of this information is captured in the high gamma range (around 70-110 Hz). Thus, gamma activity has been proposed as a robust and general indicator of local cortical function. ECoG can also reveal functional connectivity and resolve finer task-related spatial-temporal dynamics, thereby advancing our understanding of large-scale cortical processes. It has especially proven useful for advancing brain-computer interfacing (BCI) technology for decoding a user{\textquoteright}s intentions to enhance or improve communication and control. Nevertheless, human ECoG data are often hard to obtain because of the risks and limitations of the invasive procedures involved, and the need to record within the constraints of clinical settings. Still, clinical monitoring to localize epileptic foci offers a unique and valuable opportunity to collect human ECoG data. We describe our methods for collecting recording ECoG, and demonstrate how to use these signals for important real-time applications such as clinical mapping and brain-computer interfacing. Our example uses the BCI2000 software platform and the SIGFRIED method, an application for real-time mapping of brain functions. This procedure yields information that clinicians can subsequently use to guide the complex and laborious process of functional mapping by electrical stimulation. PREREQUISITES AND PLANNING: Patients with drug-resistant partial epilepsy may be candidates for resective surgery of an epileptic focus to minimize the frequency of seizures. Prior to resection, the patients undergo monitoring using subdural electrodes for two purposes: first, to localize the epileptic focus, and second, to identify nearby critical brain areas (i.e., eloquent cortex) where resection could result in long-term functional deficits. To implant electrodes, a craniotomy is performed to open the skull. Then, electrode grids and/or strips are placed on the cortex, usually beneath the dura. A typical grid has a set of 8 x 8 platinum-iridium electrodes of 4 mm diameter (2.3 mm exposed surface) embedded in silicon with an inter-electrode distance of 1cm. A strip typically contains 4 or 6 such electrodes in a single line. The locations for these grids/strips are planned by a team of neurologists and neurosurgeons, and are based on previous EEG monitoring, on a structural MRI of the patient{\textquoteright}s brain, and on relevant factors of the patient{\textquoteright}s history. Continuous recording over a period of 5-12 days serves to localize epileptic foci, and electrical stimulation via the implanted electrodes allows clinicians to map eloquent cortex. At the end of the monitoring period, explantation of the electrodes and therapeutic resection are performed together in one procedure. In addition to its primary clinical purpose, invasive monitoring also provides a unique opportunity to acquire human ECoG data for neuroscientific research. The decision to include a prospective patient in the research is based on the planned location of their electrodes, on the patient{\textquoteright}s performance scores on neuropsychological assessments, and on their informed consent, which is predicated on their understanding that participation in research is optional and is not related to their treatment. As with all research involving human subjects, the research protocol must be approved by the hospital{\textquoteright}s institutional review board. The decision to perform individual experimental tasks is made day-by-day, and is contingent on the patient{\textquoteright}s endurance and willingness to participate. Some or all of the experiments may be prevented by problems with the clinical state of the patient, such as post-operative facial swelling, temporary aphasia, frequent seizures, post-ictal fatigue and confusion, and more general pain or discomfort. At the Epilepsy Monitoring Unit at Albany Medical Center in Albany, New York, clinical monitoring is implemented around the clock using a 192-channel Nihon-Kohden Neurofax monitoring system. Research recordings are made in collaboration with the Wadsworth Center of the New York State Department of Health in Albany. Signals from the ECoG electrodes are fed simultaneously to the research and the clinical systems via splitter connectors. To ensure that the clinical and research systems do not interfere with each other, the two systems typically use separate grounds. In fact, an epidural strip of electrodes is sometimes implanted to provide a ground for the clinical system. Whether research or clinical recording system, the grounding electrode is chosen to be distant from the predicted epileptic focus and from cortical areas of interest for the research. Our research system consists of eight synchronized 16-channel g.USBamp amplifier/digitizer units (g.tec, Graz, Austria). These were chosen because they are safety-rated and FDA-approved for invasive recordings, they have a very low noise-floor in the high-frequency range in which the signals of interest are found, and they come with an SDK that allows them to be integrated with custom-written research software. In order to capture the high-gamma signal accurately, we acquire signals at 1200Hz sampling rate-considerably higher than that of the typical EEG experiment or that of many clinical monitoring systems. A built-in low-pass filter automatically prevents aliasing of signals higher than the digitizer can capture. The patient{\textquoteright}s eye gaze is tracked using a monitor with a built-in Tobii T-60 eye-tracking system (Tobii Tech., Stockholm, Sweden). Additional accessories such as joystick, bluetooth Wiimote (Nintendo Co.), data-glove (5(th) Dimension Technologies), keyboard, microphone, headphones, or video camera are connected depending on the requirements of the particular experiment. Data collection, stimulus presentation, synchronization with the different input/output accessories, and real-time analysis and visualization are accomplished using our BCI2000 software. BCI2000 is a freely available general-purpose software system for real-time biosignal data acquisition, processing and feedback. It includes an array of pre-built modules that can be flexibly configured for many different purposes, and that can be extended by researchers{\textquoteright} own code in C++, MATLAB or Python. BCI2000 consists of four modules that communicate with each other via a network-capable protocol: a Source module that handles the acquisition of brain signals from one of 19 different hardware systems from different manufacturers; a Signal Processing module that extracts relevant ECoG features and translates them into output signals; an Application module that delivers stimuli and feedback to the subject; and the Operator module that provides a graphical interface to the investigator. A number of different experiments may be conducted with any given patient. The priority of experiments will be determined by the location of the particular patient{\textquoteright}s electrodes. However, we usually begin our experimentation using the SIGFRIED (SIGnal modeling For Realtime Identification and Event Detection) mapping method, which detects and displays significant task-related activity in real time. The resulting functional map allows us to further tailor subsequent experimental protocols and may also prove as a useful starting point for traditional mapping by electrocortical stimulation (ECS). Although ECS mapping remains the gold standard for predicting the clinical outcome of resection, the process of ECS mapping is time consuming and also has other problems, such as after-discharges or seizures. Thus, a passive functional mapping technique may prove valuable in providing an initial estimate of the locus of eloquent cortex, which may then be confirmed and refined by ECS. The results from our passive SIGFRIED mapping technique have been shown to exhibit substantial concurrence with the results derived using ECS mapping. The protocol described in this paper establishes a general methodology for gathering human ECoG data, before proceeding to illustrate how experiments can be initiated using the BCI2000 software platform. Finally, as a specific example, we describe how to perform passive functional mapping using the BCI2000-based SIGFRIED system.

}, keywords = {BCI2000, brain-computer interfacing, Electrocorticography, epilepsy monitoring, functional brain mapping, issue 64, Magnetic Resonance Imaging, MRI, neuroscience, SIGFRIED}, issn = {1940-087X}, doi = {10.3791/3993}, url = {http://www.ncbi.nlm.nih.gov/pubmed/22782131}, author = {Jeremy Jeremy Hill and Disha Gupta and Peter Brunner and Gunduz, Aysegul and Adamo, Matthew A and A L Ritaccio and Gerwin Schalk} } @article {2853, title = {Treatments/Interventions for Autism @ Autism Outreach Albany Workshop www.autism-outreach.com }, year = {2012}, month = {09/2012}, author = {Disha Gupta} } @article {2852, title = {Auditory Processing and Anticipation @ The 3rd International Workshop on Advances in Electrocorticography}, year = {2011}, month = {11/2011}, author = {Disha Gupta} } @article {2841, title = {Space{\textendash}time network connectivity and cortical activations preceding spike wave discharges in human absence epilepsy: a MEG study.}, journal = {Medical and Biological Engineering and Computing}, volume = {49}, year = {2011}, month = {05/2011}, chapter = {555}, abstract = {To describe the spatial and temporal profiles of connectivity networks and sources preceding generalized spike-and-wave discharges (SWDs) in human absence epilepsy. Nonlinear associations of MEG signals and cluster indices obtained within the framework of graph theory were determined, while source localization in the frequency domain was performed in the low frequency bands with dynamic imaging of coherent sources. The results were projected on a three-dimensional surface rendering of the brain using a semi-realistic head model and MRI images obtained for each of the five patients studied. An increase in clustering and a decrease in path length preceding SWD onset and a rhythmic pattern of increasing and decreasing connectivity were seen during SWDs. Beamforming showed a consistent appearance of a low frequency frontal cortical source prior to the first generalized spikes. This source was preceded by a low frequency occipital source. The changes in the connectivity networks with the onset of SWDs suggest a pathologically predisposed state towards synchronous seizure networks with increasing connectivity from interictal to preictal and ictal state, while the occipital and frontal low frequency early preictal sources demonstrate that SWDs are not suddenly arising but gradually build up in a dynamic network.}, keywords = {Absence epilepsy, Beamforming, Connectivity, Magnetoencephalography, Nonlinear association analysis, Small world networks, Spike wave discharge}, doi = {10.1007/s11517-011-0778-3}, url = {http://www.ncbi.nlm.nih.gov/pubmed/21533620}, author = {Disha Gupta and Pauly Ossenblok and Gilles van Luijtelaar} } @article {2842, title = {Dynamic imaging of {\textquoteleft}generalized{\textquoteright} seizure activity: clinical MEG workshop @ Sleep and Epilepsy update: 12th annual international clinical symposium Kempenhaeghe}, year = {2010}, author = {Disha Gupta} } @article {2845, title = {Seizure prediction for epilepsy using a multi-stage phase synchrony based system.}, journal = {Conf Proc IEEE Eng Med Biol Soc}, volume = {2009}, year = {2009}, month = {09/2009}, pages = {25-8}, abstract = {Seizure onset prediction in epilepsy is a challenge which is under investigation using many and varied signal processing techniques. Here we present a multi-stage phase synchrony based system that brings to bear the advantages of many techniques in each substage. The 1(st) stage of the system unmixes continuous long-term (2-4 days) multichannel scalp EEG using spatially constrained Independent Component Analysis and estimates the long term significant phase synchrony dynamics of narrowband (2-8 Hz and 8-14 Hz) seizure components. It then projects multidimensional features onto a 2-D map using Neuroscale and evaluates the probability of predictive events using Gaussian Mixture Models. We show the possibility of seizure onset prediction within a prediction window of 35-65 minutes with a sensitivity of 65-100\% and specificity of 65-80\% across epileptic patients.}, keywords = {Algorithms, Artificial Intelligence, Diagnosis, Computer-Assisted, Electroencephalography, Epilepsy, Humans, Pattern Recognition, Automated, Reproducibility of Results, Sensitivity and Specificity}, issn = {1557-170X}, doi = {10.1109/IEMBS.2009.5334898}, url = {http://www.ncbi.nlm.nih.gov/pubmed/19965104}, author = {Christopher J James and Disha Gupta} } @conference {2849, title = {Phase synchronization with ICA for epileptic seizure onset prediction in the long term EEG.}, booktitle = {4th IET International Conference on Advances in Medical, Signal and Information Processing}, year = {2008}, month = {07/2008}, publisher = {IET}, organization = {IET}, address = {Santa Margherita Ligure}, abstract = {The apparently unpredictable nature of epileptic seizures can be devastating for people with epilepsy. Current medical interventions can help 75\% of patients while 25\% have to live with uncontrolled seizures. This motivates the search for a seizure prediction prototype using electroencephalograms (electrical signals that capture brain activity). The concept of phase synchrony has attracted much attention recently in the context of seizure prediction but is still in need of further study. The basis of our analysis is to track changes in synchrony in brain signals at and before seizure onset. The novel concept in our analysis is the use of unmixed signals as opposed to scalp EEG signals for phase synchrony analysis. The unmixing is performed by a Blind Source Separation technique called Independent component Analysis (ICA). ICA seeks underlying independent source signals from the EEG and it allows multivariate analysis using spatial as well as temporal information inherent to EEG signals. The present study on long-term continuous EEG data sets indicates that the concept of using phase synchronization with ICA may prove useful for predicting seizures.}, isbn = {978-0-86341-934-8}, url = {http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=4609101\&abstractAccess=no\&userType=inst}, author = {Disha Gupta and Christopher J James and William P Gray} } @conference {2847, title = {Flux-continuous schemes for solving EEG source localization problems.}, booktitle = {15th UK Conference of the Association of Computational Mechanics in Engineering 2007 }, year = {2007}, month = {07/2007}, publisher = {Civil-Comp Press, Curran Associates, Inc.}, organization = {Civil-Comp Press, Curran Associates, Inc.}, address = {Glasgow, UK}, keywords = {control volume distributed, electroencephalographic, finite element method, flux-continuous schemes, independent component analysis, Poisson{\textquoteright}s equation, source localization}, doi = {10.4203/ccp.85.13}, url = {http://www.ctresources.info/ccp/paper.html?id=4312}, author = {Mayur Pal and Disha Gupta and M G Edwards and Christopher J James} } @article {2844, title = {Narrowband vs. broadband phase synchronization analysis applied to independent components of ictal and interictal EEG.}, journal = {Conf Proc IEEE Eng Med Biol Soc}, volume = {2007}, year = {2007}, month = {08/2007}, pages = {3864-7}, abstract = {This paper presents a comparison of the use of broadband and narrow band signals for phase synchronization analysis as applied to Independent Components of ictal and interictal scalp EEG in the context of seizure onset detection and prediction. Narrow band analysis for phase synchronization is found to be better performed in the present context than the broad band signal analysis. It has been observed that the phase synchronization of Independent Components in a narrow band (particularly the Gamma band) shows a prominent trend of increasing and decreasing synchronization at seizure onset near the epileptogenic area (spatially). This information is not always found to be consistent in analysis with the raw EEG signals, which may show spurious synchronization happening due to volume conduction effects. These observations lead us to believe that tracking changes in phase synchronization of narrow band activity, on continuous data records will be of great value in the context of seizure prediction.}, keywords = {Algorithms, Electroencephalography, Humans, Predictive Value of Tests, Seizures, Signal Processing, Computer-Assisted}, issn = {1557-170X}, doi = {10.1109/IEMBS.2007.4353176}, url = {http://www.ncbi.nlm.nih.gov/pubmed/18002842}, author = {Disha Gupta and Christopher J James} } @article {2846, title = {Space-time ICA versus Ensemble ICA for ictal EEG analysis with component differentiation via Lempel-Ziv complexity.}, journal = {Conf Proc IEEE Eng Med Biol Soc}, volume = {08/2007}, year = {2007}, month = {2007}, pages = {5473-6}, abstract = {In this proof-of-principle study we analyzed intracranial electroencephalogram recordings in patients with intractable focal epilepsy. We contrast two implementations of Independent Component Analysis (ICA) - Ensemble (or spatial) ICA (E-ICA) and Space-Time ICA (ST-ICA) in separating out the ictal components underlying the measurements. In each case we assess the outputs of the ICA algorithms by means of a non-linear method known as the Lempel-Ziv (LZ) complexity. LZ complexity quantifies the complexity of a time series and is well suited to the analysis of non-stationary biomedical signals of short length. Our results show that for small numbers of intracranial recordings, standard E-ICA results in marginal improvements in the separation as measured by the LZ complexity changes. ST-ICA using just 2 recording channels both near and far from the epileptic focus result in more distinct ictal components--although at this stage there is a subjective element to the separation process for ST-ICA. Our results are promising showing that it is possible to extract meaningful information from just 2 recording electrodes through ST-ICA, even if they are not directly over the seizure focus. This work is being further expanded for seizure onset analysis.}, keywords = {Algorithms, Artificial Intelligence, Diagnosis, Computer-Assisted, Electroencephalography, Epilepsy, Humans, Pattern Recognition, Automated, Principal Component Analysis, Reproducibility of Results, Sensitivity and Specificity}, issn = {1557-170X}, doi = {10.1109/IEMBS.2007.4353584}, url = {http://www.ncbi.nlm.nih.gov/pubmed/18003250}, author = {Christopher J James and Ab{\'a}solo, Daniel and Disha Gupta} } @conference {2848, title = {De-noising epileptic EEG using ICA and phase synchrony.}, booktitle = {3rd International Conference on Advances in Medical, Signal and Information Processing, IET}, year = {2006}, month = {07/2006}, publisher = {IET, Curran Associates, Inc.}, organization = {IET, Curran Associates, Inc.}, address = {Glasgow, Scotland}, abstract = {A multi-channel recording of scalp electroencephalogram (EEG) is a non-invasive tool important for analysis and treatment of patients with epilepsy. These recordings are usually contaminated with artifacts and background activity, which may sometimes render them misleading or useless. Epileptic EEG is also useful for seizure detection, localisation and prediction. It would be useful to de-noise epileptic EEG in order to improve the efficiency of such diagnostic and prognostic procedures. The basic method of denoising a signal is through filtering, but filtering physiological signals is not trivial and highly subjective as the information is spread over different frequency bands and different measurement channels. This paper demonstrates a system for objectively de-noising epileptic EEG using Independent Component Analysis (ICA). In the standard implementation of ICA it is generally required to subjectively choose independent components (ICs) relevant to the epileptic activity; here we automate this process through the concept of phase synchronisation between ICs. In this manner de-noising the epileptic EEG with ICA becomes an objective (and automated) process.}, isbn = {978-0-86341-658-3}, doi = {10.1049/cp:20060398}, url = {http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=4225235}, author = {Disha Gupta and Christopher J James and William P Gray} } @mastersthesis {2843, title = {Advances in Epileptic Seizure Onset Prediction in the EEG with ICA and Phase Synchronization}, volume = {PhD}, school = {University of Southampton}, address = {Southampton, UK}, abstract = {Seizure onset prediction in epilepsy is a challenge which is under investigation using many and varied signal processing techniques, across the world. This research thesis contributes to the advancement of digital signal analysis of neurophysiological signals of epileptic patients. It has been studied especially in the context of epileptic seizure onset prediction, with a motivation to help epileptic patients by advancing the knowledge on the possibilities of seizure prediction and inching towards a clinically viable seizure predictor. In this work, a synchrony based multi-stage system is analyzed that brings to bear the advantages of many techniques in each substage. The 1st stage of the system unmixes and de-noises continuous long-term (2-4 days) multichannel scalp Electroencephalograms using spatially constrained Independent Component Analysis. The 2d stage estimates the long term significant phase synchrony dynamics of narrowband (2-8 Hz and 8-14 Hz) seizure components. The synchrony dynamics are assessed with a novel statistic, the PLV-d, analyzing the joint synchrony in two frequency bands of interest. The 3rd stage creates multidimensional features of these synchrony dynamics for two classes ({\textquoteleft}seizure free{\textquoteright} and {\textquoteleft}seizure predictive{\textquoteright}) which are then projected onto a 2-dimensional map using a supervised Neuroscale, a topographic projection scheme based on a Radial Basis Neural Network. The 4th stage evaluates the probability of occurrence of predictive events using Gaussian Mixture Models used in supervised and semi-supervised forms. Preliminary analysis is performed on shorter data segments and the final system is based on nine patient{\textquoteright}s long term (2-4 days each) continuous data. The training and testing for feature extraction analysis is performed on five patient datasets. The features extracted and the parameters ascertained with this analysis are then applied on the remaining four long-term datasets as a test of performance. The analysis is tested against random predictors as well. We show the possibility of seizure onset prediction (performing better than a random predictor) within a prediction window of 35-65 minutes with a sensitivity of 65-100\% and specificity of 60-100\% across the epileptic patients.}, author = {Disha Gupta}, editor = {Christopher J James} }