Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Thesis
Seizure detection using a novel multi-measurement support vector machine algorithm
Master of Science (M.S.), Drexel University
Sep 2009
DOI:
https://doi.org/10.17918/etd-3124
Abstract
Epilepsy is the second most common neurologic disorder which is characterized by recurrent and spontaneous seizures. Seizures occur unpredictably which makes everyday tasks such as driving and working extremely difficult resulting in a reduced quality of life. Although pharmaceutical treatment works well for approximately 70% of patients, 30% of epilepsy patients live with this disease unless they are eligible for surgical removal of the epileptic brain tissue responsible for initiating the seizures. Epilepsy monitoring units measure the electrical activity of the brain using electroencephalography (EEG) in an attempt to locate the epileptic brain tissue. However since seizures occur unpredictably and are generally infrequent, long recording times generate massive quantities of data which must be reviewed by neurologists to identify seizures. Automating this process using a seizure detection algorithm will ultimately save time and money, allow for superior and safer care of patients, and provide a better diagnostic tool. Although seizure detection has been well studied in the laboratory and clinic, a widely accepted algorithm has not been developed largely due to the fact that automated routines do not perform as well as a neurologist. In order to improve performance, we tested the hypothesis that multiple algorithms would work better than any single approach. Multiple algorithms were used as feature extractors and were implemented into a support vector machine (SVM) algorithm. The proposed method was optimized and tested using an animal model of epilepsy as well as human epilepsy data obtained from Hahnemann University Hospital. The results of this analysis showed that the multi-measurement SVM algorithm performed better (93.3% sensitivity, 91.7% positive predictive value) than any single measurement. Additionally, the SVM algorithm performed better than the commercially available XLTEK algorithm by obtaining a sensitivity of 98.9% and a positive predictive value of 25.5%.
Metrics
26 File views/ downloads
38 Record Views
Details
- Title
- Seizure detection using a novel multi-measurement support vector machine algorithm
- Creators
- Kevin J. Freedman - DU
- Contributors
- Karen Anne Moxon (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Master of Science (M.S.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Resource Type
- Thesis
- Language
- English
- Academic Unit
- School of Biomedical Engineering, Science, and Health Systems (1997-2026); Drexel University
- Other Identifier
- 3124; 991014632594404721