Mackenzie, Rebecca
(2024)
A Compact Machine-Learning Based Diagnostic Utility for Seizure Detection and Localization.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Seizures are episodes of abnormally excessive or synchronous electrical activity of localized populations of neurons in the brain. Epilepsy is a condition in which these seizures are repetitive. Some half a million people do not respond to drug therapy, and surgical intervention is necessary. Multi-electrode EEG recording is an important diagnostic tool for pre-surgical planning in these cases. While EEGs have relatively good temporal resolution, they have poor spatial resolution. Therefore, the onset and localization of seizure activity can be difficult to determine by direct observation of the EEG record. If those EEG channels that show the earliest onset of seizure-like activity can be determined, then those channels, corresponding to locations on the cerebral cortex, can indicate sites for the surgical correction of epileptogenic foci. To assist the clinician in determining which channels correspond to the onset of seizure activity, machine-learning tool have shown promise. However, computer-based diagnostic systems that are portable and accurate are not readily available for use by clinicians, especially in underserved areas. Also, the numerous computer-based EEG analysis utilities that have been devised to detect seizures remain limited in one or more respects: 1. a lack of sufficient spatial resolution of the EEG signal to indicate activity at a specific electrode, 2. an inability to resolve both temporal and spatial information for the same EEG signal, 3. the lack of ability to store previous analysis sets, so the system can refine its learning accuracy using new data, and 4. the ability to be quickly and accurately run on a portable computer system for use in underserved, rural, or remote
v
geographical regions. Previously, I demonstrated that a relatively simple recurrent neural network was capable of seizure detection and localization from an EEG record. To improve on this system and address the four limitations listed above, I have developed a compact machine-learning utility based on my original system but with significant innovations. This system has shown its effectiveness in determining the onset and cortical localization of seizure activity from high-resolution EEG records.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
3 June 2024 |
| Date Type: |
Publication |
| Defense Date: |
19 March 2024 |
| Approval Date: |
3 June 2024 |
| Submission Date: |
4 April 2024 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
110 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Swanson School of Engineering > Bioengineering |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Epilepsy seizures machine-learning localization |
| Date Deposited: |
03 Jun 2024 14:41 |
| Last Modified: |
03 Jun 2024 14:41 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/46019 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}