Seizure and Waveform Analysis and Detection
2.1) Low Power Interictal Detection Algorithm to Facilitate Long Term and Wireless AEEG Monitoring
Alexander Casson, David Yates, Esther Rodriguez-Villegas
Circuits and systems research group, Department of Electrical and Electronic Engineering, Imperial College London
Traditional seizure and interictal detection techniques aim to quantify the amount of activity present. To do this they must have a high sensitivity (to correctly detect all of the events) and few false detections (high specificity). These requirements are difficult to fulfil simultaneously. An alternative approach, termed data selection, is presented here. This method is a different view on the detection problem: A detection procedure is used to select which sections of data are saved—a fixed amount on either side of an automated detection—and which are discarded. A high sensitivity is still required, but false detections are not as significant as they are rejected by a human interpreter in the same way as background signals are when a standard continuous EEG trace is analysed. The method only allows data reduction, not event quantification, but this is still significant in reducing the analysis time required and in facilitating wireless ambulatory EEG units. Ordinarily there is too much raw EEG data to transmit without compromising battery life and so online, low power, data reduction is required. The data selection algorithm’s tolerance to false detections simplifies the algorithm design making it very suitable for this low power implementation. The data selection method presented here is based upon the Continuous Wavelet Transform and can offer a 50% reduction in the amount of data to be transmitted whilst correctly recording 95% of expert marked interictal events. All of the algorithm blocks are also suitable for ultra low power VLSI implementation.
2.2) Improved Temporal Lobe Epileptic Event Detection through Inclusion of Cardiac Fluctuations.
Thomas Bermudez (1), David Lowe (1), Anne-Marie Arlaud-Lamborelle (2)
(1) Aston University, Birmingham, United Kingdom; (2) Centre Hospitalier Henri Duffaut, Avignon, France
Due to the activation of the central autonomic network, neurovegetative signs are commonly observed during epileptic seizures either as the primary seizure manifestation or as an accompaniment to other seizure symptoms. Autonomic changes (cardiovascular changes, respiratory manifestations, gastrointestinal symptoms, cutaneous manifestations, pupillary symptoms and genital and sexual manifestations) give valuable information on the topographic origin of the epileptiform discharge. The cardiovascular autonomic symptoms induced by temporal lobe epilepsy seizures mainly consist of tachycardia often with an increase of blood pressure; rarely they consist of bradycardia, dramatic decrease of blood pressure; and irregular pulse. The role of epileptiform discharges localised to temporal lobe structures in the induction of neurovegetative heart arrhytmias has been demonstrated: on stimulation of the human left insular cortex bradycardia and depressor responses were more frequently obtained than tachycardia and pressor effects whereas the converse applied for the right insular cortex. The analysis of the R-R intervals variability from electrocardiographic recordings has been used as a measure of autonomic control over the cardiovascular system to detect periods of increased seizure likelihood. We consider the Brain-Heart system as a coupled system in which bioenergetic processes in the brain have an autonomic influence on the heart. We specifically investigate temporal lobe epilepsy and its correlation to cardiac arrhythmias to develop a probabilistic model fusion approach applied to simultaneously recorded EEG and ECG data for both ictal and interictal episodes and provide evidence that epileptic event detection is improved through the inclusion of a probabilistic description of the cardiac fluctuations.
2.3) Automatic Detection and Ictal Pattern Recognition of Epileptic Seizures in Long-Term Human EEG
Ralph Meier (1,2), Andreas Schulze-Bonhage (2,1), Ad Aertsen (1,3)
(1) Bernstein Center for Computational Neuroscience, Freiburg; (2) Center for Epilepsy, Freiburg; (3) Faculty of Biology, Albert-Ludwigs-University, Freiburg
Epileptic seizures are reflected in human EEG by multiple ictal patterns. Recently, the prospect of warning and potential intervention systems based on the reliably, early detection of ictal EEG patterns have attracted increasing interest. Moreover, since the workload involved in the detection of seizures by human experts is quite formidable, several attempts have been made to develop automatic seizure detection systems. Here we present a novel procedure for generic, online and real-time automatic detection of multi-morphological ictal-patterns in the human long-term EEG and its validation in continuous, routine clinical EEG recordings from 57 Patients with a duration of approximately 43h. We analysed 91 seizures representing the 6 most common ictal morpgologies (Alpha, Beta, Theta and Delta- rhythmic activity, Amplitude depression and Polyspikes). We found that taking the seizure morphology into account plays a crucial role in optimization of the detection performance. Moreover, besides enabling a reliable (mean false alarm rate <0.5/h, for specific ictal morphologies <0.25/h), early and accurate detection (average correct detection rate >96%) within the first few seconds (average latency of approx. 2s) of ictal patterns in the EEG, this procedure facilitates the automatic categorization of the prevalent seizure morphologies without the neccessity to adapt the proposed system to specific patients. Acknowledgements: We thank Armin Brandt, Carolin Gierschner and Heike Dittrich from the Freiburg Epilepsy Center. Partial funding for this research was supplied by the Committee for Research, University Clinics, Freiburg and the German Federal Ministry for Education and Research (BMBF, grant 01GQ0420 to BCCN Freiburg).
2.4) Analysis of the Dynamics of Human Epileptic Seizures from Scalp EEG
Nadia Mammone (1), Umberto Aguglia (2), Maurizio Fiasché (1), Fabio La Foresta (1), Emilio Le Piane (2), Francesco Carlo Morabito (1)
(1) Neurolab, DIMET, University of Reggio Calabria; (2) Centro Regionale Epilessie, Università Magna Graecia di Catanzaro, Presidio Riuniti, Reggio Cal.
Results in literature show that the convergence of the STLmax time series, extracted from intracranial EEG recorded from patients affected by intractable temporal lobe epilepsy, is linked to the seizure onset: when the convergence of STLmax profiles of critical electrode sites reaches a critical level, a seizure is likely to occur. Moreover, the behaviour of this convergence over time allows for the automatic detection of the electrodes involved in the process leading to the seizure. This prediction technique is called ASPA and was presented by Iasemidis et al. in 2003. In order to analyse the predictability of seizures from scalp EEG, this prediction technique was here implemented and tested over three scalp EEG recordings: One from a patient affected by partial frontal lobe epilepsy (patient A) and two from a patient affected by absence seizures (patient B). The algorithm exploits the first seizure in each recording for selecting the groups of critical electrodes to be monitored, this process takes 10min after the seizure onset. After the initialization, the technique succeeded in issuing a warning before every seizure, with an average horizon of 5.43min, that is a good result since the average monitoring time was 6.17min. The technique also automatically detected as critical the electrodes in the focal area, for patient A, and in the frontal area, for patient B. In conclusion, ASPA seems to be able to detect changes also in the dynamics of scalp EEG as well as to infer information about the critical area, even for absence seizures.
2.5) Seizure Detection Enhanced by Seizure Prediction
Levin Kuhlmann, Iven Mareels
Department of Electrical and Electronic Engineering, The University of Melbourne
Seizure detection involving seizure probability estimation has been applied to intracranial data in a patient non-specific manner (Clinical Neurophysiology 116 (2005) 2460–2472). This method has been shown to give performance levels acceptable for clinical use. Here this method is combined with a seizure prediction method based on phase synchronization (Physical Review E 67 (2003) 021912), in order to boost its detection performance. The seizure predictor is tuned to have a very low false positive rate (FPR), and hence low sensitivity. Despite this low sensitivity, any prediction is likely to lead to a seizure because of the low false positive rate. Thus when a prediction signal is given one can temporarily lower the seizure detection probability threshold, in order to increase the likelihood of detecting a seizure. The feasibility of this method was tested on the contest data for the Freiburg Seizure Prediction Workshop. In order for prediction to enhance detection, the predictor needs to predict seizures that the detector cannot detect, and this was difficult to do given the low sensitivity of the seizure predictor and the small number of seizures in the contest data. The feasibility of this method will need to be evaluated on a larger data set.
2.6) Multiscale Electrophysiology in Human Epileptogenic Brain: Microseizures, DC-fluctuations, and High Frequency Oscillations
Matt Stead (1), Sanqing Hu (1), Andrew Gardner (2), Brian Litt (2), Kendall Lee (1), Greg Worrell (1)
(1) Mayo Clinic, Department of Neurology, Rochester, MN 55905; (2) University of Pennsylvania, Department of Neurology, Philadelphia, PA, 19104
RATIONALE: Human brain oscillations span a wide range of spatiotemporal scales. The spatial organization of neuronal assemblies range from small neuronal clusters to centimeter scale networks. Similarly, the frequencies span a wide range from DC to high frequency oscillations. METHODS: We studied 10 patients with hybrid electrodes containing microwires and macroelectrodes. The EEG was acquired using a DC capable broadband amplifier (Neuralynx Inc.). RESULTS: Broadband recordings demonstrated oscillations extending from DC to 700 Hz. Slow oscillations and DC fluctuations were often spatially diffuse, but also occurred in sub-millimeter regions within the seizure onset zone. Fast oscillations (>80 Hz) were primarily localized, and most prominent within the seizure onset zone. In the seizure onset zone frequent sub-millimeter domain seizures were recorded (microseizures) on the microwires, but not apparent on clinical macroelectrodes. The microseizure activity was spatially and temporally correlated with large-scale clinical seizure activity. CONCLUSIONS: Multiscale EEG recordings demonstrate oscillations and seizures occur over a range of spatiotemporal scales. The microwires demonstrate independent microdomains of seizure activity occurring throughout the epileptogenic zone. Remarkably, microseizures are not detected with standard macroelectrodes. We propose that the generation of focal seizures may occur by the coalescence of microseizure islands, and only become apparent on macroelectrodes after sufficient tissue has been recruited.
2.7) Spread of Ictal Activity in Focal Epilepsy of Frontal and Temporal Origin
Katrin Götz-Trabert, Christoph Hauck, Kathrin Wagner, Susanne Fauser, Andreas Schulze-Bonhage
Epilepsy Centre University of Freiburg
Objective: Latencies between seizure onset, propagation of ictal activity, and initial clinical symptoms and signs are critically important for the successful implementation of detection-based intervention systems in the treatment of epilepsy. This study analyzes intracranial EEG-recordings for temporal characteristics of ictal spread and its dependence on focus localization. Methods: Intracerebral EEG recordings of 215 seizures from 43 patients with pharmacoresistant focal epilepsy were evaluated based on site of first propagation, latencies between EEG seizure onset, early propagation, and clinical seizure onset. Seizure onset was mesial temporal in 15 patients, neocortical temporal in 15 patients, and frontal in 13 patients. Results: Periods during which ictal activity remained confined to the seizure onset area were significantly different between the patient groups. Median latencies between electrographic seizure onset and early propagation were significantly longer for patients with mesial temporal (10 s) as compared to neocortical temporal (5 s) and frontal seizure focus (2 s; p<0.01). Concordantly, median latencies to onset of clinical symptomatology were significantly longer for patients with mesial temporal (19 s) as compared to neocortical temporal (17 s) and frontal seizure focus (6 s; p<0.01). Conclusions: The speed of propagation of ictal activity and the latencies until initial clinical seizure symptoms differ significantly depending on focus localization. Extended spread often occurred within the time window during which current detection systems operate. This suggests that inclusion criteria of patients suitable for testing the efficacy of detection-based seizure intervention strategies should be based on focus localization and patient-individual propagation patterns.
2.8) Interictal High Frequency Oscillations (>200 Hz) Recorded with Intracranial Depth Macroelectrodes: A Marker of Mesial Temporal Lobe Epilepsy?
Benoît Crepon (1,2), Vincent Navarro (1,2), Claude Adam (1,2), Michel Baulac (2), Michel Le Van Quyen (1)
(1) Laboratoire de Neurosciences Cognitives et Imagerie Cérébrale, LENA, CNRS UPR 640, Hôpital de la Pitié-Salpêtrière, Paris, France; (2) Unité d’épileptologie, Hôpital de la Pitié-Salpêtrière
Interictal surrogate markers may provide an accurate mean of localizing the epileptogenic region during the presurgical evaluation of refractory partial epilepsies. In this context, microelectrode recordings from patients with mesial temporal lobe epilepsy (MTLE) have revealed that interictal high-frequency oscillations (HFOs) from 200 to 500Hz, termed fast ripples, are specific of ictogenic structures. Our aim is to use conventional intracranial EEG recordings to characterize these pathological HFOs. Sixteen patients with intractable partial seizures undergoing investigation with intracranial macroelectrodes were recruited prospectively and consecutively. 52 minutes on average of interictal EEG sampled at 1024Hz was analyzed for each patient. HFOs (>200Hz) were detected according to an original semi automatic strategy that consisted in a full automated high sensibility detection followed by a visual validation. This selection was assisted by a wavelet decomposition based time frequency map. Our strategy permits a high sensibility and specificity detection of HFOs. They had a mean frequency of 242 ± 40Hz, were of short duration (10ms) and low amplitude (16µV). HFOs were mostly nested with a sharp wave and were recorded by one or very few nearby contacts. HFOs were recorded only for each of the nine patients suffering from MTLE, especially in the ictogenic zone. Interictal HFOs > 200Hz can be recorded from conventional depth macroelectrodes. They were recorded specifically in the ictogenic zone in MTLE, suggesting that they could have valuable diagnostic utility for routine invasive presurgical localization of these epileptic foci.