Medically intractable epilepsy is a disabling and destructive neurological disorder affecting about 0.25% of the general population. In selected patients with medication-resistant epilepsy, surgical resection of epileptogenic brain tissue has been a highly effective form of therapy. Prior to surgery, candidates for resective therapy are often implanted with intracranial depth, strip, or grid electrodes over the surface of the brain or within the substance of the brain. These electrodes allow recording subdural electroencephalograms (SEEGs) during seizures. The recorded SEEGs are then carefully examined to localize epileptic foci inferred by observing the earliest ictal patterns and the electrode sites at which such patterns arise. This data examination is critical in formulating treatment plans and surgical approach; however, visual identification of these SEEG patterns can be difficult because of the obscuring effect of various non-seizure related ongoing activities that are admixed with the early ictal activity. It is highly desirable to filter out background activities from SEEGs and unveil the early ictal activity to localize epileptic foci more accurately and reliably. This project focuses on segregating the early ictal activity and background activities in SEEGs using advanced digital signal processing techniques. Three complementary approaches will be utilized: 1) wavelet transforms and wavelet packet analysis, 2) time-frequency analysis and synthesis, and 3) adaptive filtering using recurrent artificial neural networks. These approaches are capable of handling nonstationary, nonlinear, and low signal-to-noise ratio SEEGs. The results of filtering and analysis will be evaluated by data simulation as well as examination and comparison of previously archived patient records.
| Sclabassi, R J; Sonmez, M; Sun, M (2001) EEG source localization: a neural network approach. Neurol Res 23:457-64 |
| Sun, M; Scheuer, M L; Sclabassi, R J (2001) Extraction and analysis of early ictal activity in subdural electroencephalogram. Ann Biomed Eng 29:878-86 |
| Sun, M; Sclabassi, R J (2000) The forward EEG solutions can be computed using artificial neural networks. IEEE Trans Biomed Eng 47:1044-50 |
