The Diagnosis of Psychogenic Non-epileptic Seizures using Machine Learning
- 1. Centre for Biomedical Engineering, Department of Mechanical Engineering Sciences, University of Surrey, UK
- 2. Atkinson Morley Wing, St George's Hospital, London
- 3. Department of Computer Science, University of Surrey, UK
Description
This pilot study explored the use of machine learning as a tool to replace the current gold standard method of the differential diagnosis of psychogenic non-epileptic seizures (PNES) and epilepsy. This study extracted ten temporal and spectral features from the electroencephalograms of seven subjects with PNES and seven subjects with epilepsy. Seven machine learning models were used for classification, which were tested using leave-one-subject-out cross-validation. The results were promising, with four of these models achieving average accuracies and F1-scores of over 80%, the best of which was linear discriminant analysis with 88% for both measures. In conclusion, this study has shown that machine learning is a possible tool for the differential diagnosis of PNES and epilepsy, but with a limited sample size further investigation is required.
Files
new_horizons_conference.pdf
Files
(2.3 MB)
Name | Size | Download all |
---|---|---|
md5:3529735f90f97ecd3d25b7595b0805cb
|
2.3 MB | Preview Download |
Additional details
References
- Brown RJ, Reuber M. Clin. Psychol. Rev. 2016; 47:55-70.
- Benbadis SR, Hauser WA. Seizure. 2000; 9:280-281.
- Bayly J et al. Epilepsia. 2013; 54(8):1402-1408.
- Reuber M et al. Neurology. 2002; 58(3): 493-495.
- Devinsky O, Gazzola D, LaFrance WC. Nat. Rev. Neurol. 2011; 7:210–220.