Validation of a real time seizure detection algorithm and control towards a closed loop epilepsy prosthesis

Arjun Jaitli, Purdue University

Abstract

Epilepsy is one of the most dominant neurological disorders affecting approximately 2% of the world's total population. Only about 70% of the patients suffering from chronic epilepsy are responsive to pharmacological drug treatment. Closed-loop neurostimulation devices that stimulate the seizure focus in the brain electrically to suppress seizures have shown great promises in the past. Seizure detection algorithms play an important role in facilitating responsive therapeutic intervention that increases its efficacy by improving its temporal and spatial specificity. The current study proposes a cascaded two stage seizure detection algorithm that is computationally efficient resulting in a low power hardware implementation. Additionally, a constant current stimulator prototype designed on a hardware platform for responsive therapeutic intervention has been included in the appendix section. Lastly, the study also includes testing of an event-based seizure detection algorithm implemented on hardware by our group in the past on data obtained from rats that were having spontaneous seizures. Unlike the traditional algorithms the proposed cascaded two stage algorithm does not require a `training' phase from individual to individual. It detects seizures by extracting features that show a distinct pattern at the electrographic seizure onset. The seizure detection algorithm was tested on human EEG data obtained from the Freiburg online database and performed with an overall specificity and sensitivity of 99.69% and 90.28% respectively with an average detection delay of 7.8 s [2.3, 16.5] from electrographic onset. The event-based seizure detection algorithm was validated on spontaneous seizure data obtained from chronically epileptic rats and performed with a sensitivity and specificity of 96% and 99.69% respectively with an average detection delay of approximately 4.2 s [1, 16]. The proposed techniques have been shown to be computationally efficient resulting in a low-power hardware implementation for the development of an implantable closed loop epilepsy prosthesis device that can detect and control epileptic seizures.

Degree

M.S.B.M.E.

Advisors

Irazoqui, Purdue University.

Subject Area

Engineering

COinS