Author Date

2024-06-21

Degree Name

BS

Department

Statistics

College

Physical and Mathematical Sciences

Defense Date

2024-06-18

Publication Date

2024-06-21

First Faculty Advisor

Micah R. Shepherd

First Faculty Reader

R. Ryley Parrish

Honors Coordinator

Del Scott

Keywords

Seizure, Signal Processing, Machine Learning, Neuroscience

Abstract

Status Epilepticus (SE) is a dangerous type of seizure that is difficult to treat and can have permanent or fatal consequences. Developing methods to properly process and classify SE in Local Field Potential (LFP) signals are important steps towards being able to predict SE in clinic and save lives. This thesis explores methods for the seizure data processing LFP data with the goal of gaining a deeper understanding of the effect of brain region and tissue preparation paradigm on power output in specific frequency ranges. The brain regions compared are the neocortex and hippocampus, and the preparation paradigms are 4AP and 0 Mg2+. This thesis also explores the use of statistical features in tree-based models for classifying SE-like behavior. A random-forest model was fit and tested on both intra-trace classification and inter-trace classification, with 99.58% and 64.97% accuracy, respectively, suggesting that other models may be better suited for this classification task.

Download

Share

COinS