Date of Award

8-2018

Degree Type

Thesis

Degree Name

Engineering (M.S.E.)

Department

School of Engineering

First Advisor

Robert Bossemeyer

Second Advisor

Samhita Rhodes

Third Advisor

Paul Fishback

Academic Year

2017/2018

Abstract

The objective of this thesis was to examine the ability of the Autoregressive Model Residual Modulation (ARRm) method to identify the Seizure Onset Zone (SOZ) in intracranial electroencephalogram (iEEG) of patients with refractory epilepsy. Patients who have not become seizure free after multiple trials of antiepileptic drugs (AEDs) may seek treatment through epilepsy surgery. Cortical electrodes are implanted directly on the cerebral cortex, then iEEG is collected. A specialized neurologist reviews the iEEG, then in consultation with the neurosurgeon, the SOZ is determined and areas of the brain may be chosen for resection. The success rate of epilepsy surgery varies, so it is apparent that identifying exactly where to resect epileptic tissue is still very challenging.

In recent research, High Frequency Oscillations (HFOs) in iEEG have shown strong relations to epileptic tissue. Automated HFO detection methods have been developed, but most involve analysis in the frequency domain and are computationally expensive. The ARRm method is implemented in the time domain and has potential to be implemented for real -time analysis. AR modeling is used to predict the iEEG and should not be able to accurately model highly nonharmonic events such as HFOs. Using a coefficient of variation involving the residuals of the model (ARRm value), interpretation of results showed that high ARRm values also corresponded to channel locations with high HFO counts, which included channels of interest identified by the epileptologist. Statistically significant (p<0.01) correlations were drawn between the ARRm value and HFO counts on channels of interest. Examination of the AR model residual during epileptogenic events revealed that the residual was highest when spikes and/or Fast Ripples occurred. These findings suggest that significantly correlated channels may indicate the presence of fast ripples or spikes occurring with fast ripples in the signal.

Share

COinS