Rajeev Agarwal

Activity‐dependent Gene Expression Correlates with Interictal Spiking in Human Neocortical Epilepsy

Summary:  Interictal spikes are hallmarks of epileptic neocortex that are used commonly in both EEG and electrocorticography (ECoG) to localize epileptic brain regions. Despite their prevalence, the exact relationship between interictal spiking and the molecular pathways that drive the production and propagation of seizures is not known. We have recently identified a common group of genes induced in human epileptic foci, including EGR1, EGR2, c‐fos, and MKP‐3. We found that the expression levels of these genes correlate precisely with the frequency of interictal activity and can thus serve as markers of epileptic activity. Here, we explore this further by comparing the expression of these genes within human epileptic neocortex to both ictal and specific electrical parameters of interictal spiking from subdural recordings prior to surgical resection in order to determine the electrical properties of the human neocortex that correlate best to the expression of these genes. Seizure frequency as well as quantitative electrophysiological parameters of interictal spikes including frequency, amplitude, duration, and area were calculated at each electrode channel and compared to quantitative real‐time RT‐PCR measurements of four activity‐dependent genes (c‐fos, EGR1, EGR2, and MKP‐3) in the underlying neocortical tissue. Local neocortical regions of seizure onset had consistently higher spike firing frequencies and higher spike amplitudes compared to nearby “control” cortex. In contrast, spike duration was not significantly different between these two areas. There was no relationship observed between seizure frequency and the expression levels of activity‐dependent genes for the patients examined in this study. However, within each patient, there were highly significant correlations between the expression of three of these genes (c‐fos, EGR1, EGR2) and the frequency, amplitude, and total area of the interictal spikes at individual electrodes. We conclude that interictal spiking is closely associated with the expression of a group of activity‐dependent transcription factors in neocortical human epilepsy. Since there was little correlation between gene expression and seizure frequency, our results suggest that interictal spiking is a stronger driving force behind these activity‐dependent gene changes and may thus participate in the development and maintenance of the abnormal neuronal hyperactivity seen in human epileptic neocortex.

Model-Based Seizure Detection for Intracranial EEG Recordings

This paper presents a novel model-based patient-specific method for automatic detection of seizures in the intracranial EEG recordings. The proposed method overcomes the complexities in the practical implementation of the patient-specific approach of seizure detection. The method builds a seizure model (set of basis functions) for a priori known seizure (the template seizure pattern), and uses the statistically optimal null filters as a building block for the detection of similar seizures. The process of modeling the template seizure is fully automatic. Overall, the detection method involves the segmentation of the template seizure pattern, rejection of the redundant and noisy segments, extraction of features from the segments to generate a set of models, selection of the best seizure model, and training of the classifier. The trained classifier is used to detect similar seizures in the remaining data. The resulting seizure detection method was evaluated on a total of 304 h of single-channel depth EEG recordings from 14 patients. The system performance is further compared to the Qu-Gotman patient-specific system using the same data. A significant improvement in the proposed system, in terms of specificity, is observed over the compared method.

Compression of long-term EEG using power spectral density

We propose to use the features based on power spectral density as a descriptor of the EEG in the compression of the long-term intensive care unit EEG to obtain the temporal evolution of the recurrent patterns. Sleep EEG is used as a baseline since the sleep stages can be mapped to recurrent patterns in the background EEG. Our results indicate that the spectral features provide a better classification of the sleep EEG and assist in a better formation of homogenous clusters compared to the results obtained with the previously used features. The average overall agreement compared against manual scoring of seven sleep EEG records is 68.5%. It is an improvement compared to 62.7% obtained with the previously used features. Although our results for computer classification use only the EEG information from one frontal and one occipital channel, they are similar to the manual classification of sleep EEG, which is based on additional information.

Morphology-Based Automatic Seizure Detector for Intracerebral EEG Recordings

In this paper, a new seizure detection system aimed at assisting in a rapid review of prolonged intracerebral EEG recordings is described. It is based on quantifying the sharpness of the waveform, one of the most important electrographic EEG features utilized by experts for an accurate and reliable identification of a seizure. The waveform morphology is characterized by a measure of sharpness as defined by the slope of the half-waves. A train of abnormally sharp waves resulting from subsequent filtering are used to identify seizures. The method was optimized using 145 h of single-channel depth EEG from seven patients, and tested on another 158 h of single-channel depth EEG from another seven patients. Additionally, 725 h of depth EEG from 21 patients was utilized to assess the system performance in a multichannel configuration. Single-channel test data resulted in a sensitivity of 87% and a specificity of 71%. The multichannel test data reported a sensitivity of 81% and a specificity of 58.9%. The new system detected a wide range of seizure patterns that included rhythmic and nonrhythmic seizures of varying length, including those missed by the experts. We also compare the proposed system with a popular commercial system.

Automatic Detection of Micro-Arousals

In patients suffering from various sleep disorders and some elderly patients, sleep is disturbed with frequent but brief arousal. These events do not cause behavioral awakening, but can lead to excessive day time sleepiness. These brief arousals or microarousals (MAs) can be identified on a standard polysomnogram as a transient abrupt change of frequency, typically in the alpha and extended beta (16-40 Hz) bands. In this paper, we present a novel method to automatically detect MAs. The method is based on using the ideas of segmentation, spectral feature extraction and the identification of EEG epochs containing MA with statistical methods and decisional rules. Full-night EEG recordings from two patients are used to present some initial performance results. For this analysis, the MA events are independently scored by three experienced sleep experts. Results show the method to be promising; however, due to the large inter-scorer variations it may be necessary to tailor the detection threshold to address the varying scorer preferences (address the sensitivity/specificity tradeoffs)

A novel dual-stage classifier for automatic detection of epileptic seizures

In long-term monitoring of electroencephalogram (EEG) for epilepsy, it is crucial for the seizure detection systems to have high sensitivity and low false detections to reduce uninteresting and redundant data that may be stored for review by the medical experts. However, a large number of features and the complex decision boundaries for classification of seizures eventually lead to a trade-off between sensitivity and false detection rate (FDR). Thus, no single classifier can fulfill the requirements of high sensitivity with a low FDR and at the same time be a computationally efficient system suitable for real-time application. We present a novel, simple, computationally efficient seizure detection system to enhance the sensitivity with a low FDR by proposing a dual-stage classifier. This overall system consists of a pre-processing unit, a feature extraction unit and a novel dual-stage classifier. The first stage of the proposed classifier detects all true seizures, but also many false patterns, whereas the second stage of the proposed classifier minimizes false detections by rejecting patterns that may be artifacts. The performance of the novel seizure detection system has been evaluated on 300 hours of single-channel depth electroencephalogram (SEEG) recordings obtained from fifteen patients. An overall improvement has been observed in terms of sensitivity, specificity and FDR.

A novel unsupervised spike sorting algorithm for intracranial EEG

This paper presents a novel, unsupervised spike classification algorithm for intracranial EEG. The method combines template matching and principal component analysis (PCA) for building a dynamic patient-specific codebook without a priori knowledge of the spike waveforms. The problem of misclassification due to overlapping classes is resolved by identifying similar classes in the codebook using hierarchical clustering. Cluster quality is visually assessed by projecting inter- and intra-clusters onto a 3D plot. Intracranial EEG from 5 patients was utilized to optimize the algorithm. The resulting codebook retains 82.1% of the detected spikes in non-overlapping and disjoint clusters. Initial results suggest a definite role of this method for both rapid review and quantitation of interictal spikes that could enhance both clinical treatment and research studies on epileptic patients.

A new improved model-based seizure detection using statistically optimal null filter

A patient-specific model-based seizure detection method using statistically optimal null filters (SONF) has been recently proposed to aid the review of long-term EEG [1, 2]. The method relies on the model of a priori known seizure (template pattern) for subsequent detection of similar seizures. Artifacts, non-epileptic EEG rhythms, and at times modeling errors lead to increased false or missed detections. In this paper, we present a new improved model-based seizure detection that introduces a pre-processing block for artifact rejection, an adaptive technique of modeling the template patterns, and a new evolution-based classifier. The proposed classifier tracks the temporal evolution of seizure to improve the classification accuracy. With the help of simulated EEG, we illustrate the significance and need for these modifications. Further, performance of the complete algorithm is tested on single channel depth EEG of seven patients, and compared with the previous approaches. In terms of sensitivity and specificity, the proposed method resulted in 84% and 100%, method of [1] 65% and 84%, and method of [2], 84% and 90% respectively. An overall performance improvement is seen as enhanced detection sensitivity and reduced false positives. This is preliminary result on seven patient data.

Detection of epileptic seizures in stereo-EEG using frequency-weighted energy

This paper proposes a new algorithm for seizure detection based on the evolution-like characteristics of a seizure. Most of the existing algorithms for automatic detection of the epileptic seizures in electroencephalograms (EEG) rely upon some pre-defined/patient-tunable detection threshold to classify the data as normal or abnormal. In this paper, we present a method for seizure detection in stereoencephalograms (SEEG) using frequency-weighted energy. The method does not require a threshold or any a priori information about the seizure for its detection. The method is gradient-based and any activity that exceeds the minimum duration satisfying our criteria is considered as a potential seizure activity. The performance of the algorithm is evaluated on 100 hours of single channel SEEG data obtained from five different patients. An overall sensitivity of 96.6% and a false detection rate of 0.21/h is obtained on the complete data.

LMS-optimal notch filters with improved transient performance

The use of a pole-contraction factor (/spl alpha/) with a value close to unity in the Constrained Notch Filters (CNFs) causes excessively long transient response. For practically useful values of /spl alpha/, the transient duration is reduced, on the other hand however, the error in transparency as measured by the MSE is increased. In this paper, we increase the order of CNFs by a strategic pole/zero placement to address this tradeoff. Two approaches of finding these pole/zero locations are shown. One of the methods is optimal in the LMS sense, while the other, though suboptimal, yields a closed-form solution. The resulting filters show a significant improvement in the transient duration.