Brian Litt

Line length: an efficient feature for seizure onset detection

A signal feature with low computational burden is presented as an efficient tool for seizure onset detection. The feature was evaluated over a total of. 1,215 hours of intracranial EEG signal from 10 patients. Results confirmed this feature as being useful for seizure onset detection yielding an average delay of 4.1 seconds, 0.051 false positives per hour, and one false negative on a subclinical seizure out of 111 seizures analyzed of which 23 were subclinical.

Feature parameter optimization for seizure detection/prediction

When dealing with seizure detection/prediction problems, there are three main performance metrics that must be optimized: false positive rate, false negative rate, detection delay or, if the problem is seizure prediction, it is desirable to obtain the greatest prediction time achievable. Tuning specific extracted features to individual patients can lead to improved results. The processing window length is also an important parameter whose optimization may significantly affect performance. In this study we propose an approach for selecting the window length for the particular detection/prediction problem. This approach is applicable to other feature parameters suitable for tuning or optimization.

A genetic approach to selecting the optimal feature for epileptic seizure prediction

The objective of this study is to (1) develop and apply efficient algorithms to simultaneous intracranial electroencephalographic signals recorded from multiple implanted electrode sites to evaluate the spatial and temporal behavior of seizure precursors and (2) to demonstrate the utility of multiple feature and channel synergy for predicting epileptic seizures in patients with mesial temporal lobe epilepsy. Short-term seizure precursors within a 10-minute time period are investigated. The method consists of preprocessing, processing, feature selection, classification, and validation steps. The preprocessing step removes extraneous data and captures the salient signal attributes while maintaining the integrity of the signal. Processing is a three-step approach that includes first-level features extracted from the raw data, second-level features extracted from first level features, and third-level features extracted from second-level features. A genetic algorithm selects the optimal features off-line from a preselected group of features to serve as the input to the classifier.

Fractal dimension detects seizure onset in mesial temporal lobe epilepsy

The fractal dimension (FD) of a waveform is presented as a tool for detecting electrographic and predicting clinical onset of seizures (Sz) in 5 patients with mesial temporal lobe epilepsy (MTLE). Sixty-seven 1-hour intracranial EEG (IEEG) records, 25 containing Szs and 42 randomly chosen baselines /spl ges/6 hrs from Szs, were analyzed from 5 patients during presurgical evaluation. Szs were detected an average of 9.8 seconds after EEG onset and 52 seconds prior to clinical onset. There were no false positive or false negative detections. It is concluded that FD of the IEEG waveform is a promising method for real-time detection of electrographic and prediction of clinical onset of epileptic Szs early enough to warn of impending events and potentially to trigger abortive therapy.

Median-based filtering methods for EEG seizure detection

Median filters can be used to detect seizures, however their response times vary widely with the chosen algorithmic implementations. The computational effort of a brute-force running median, a quick running median, and a novel modified running median filter is investigated. The modified running median was found to demand low linear-order effort, while preserving properties of the true median filters.