Article Text

Download PDFPDF

B165 Comparison of machine learning algorithms in predicting EEG epileptic seizure during anaesthesia
  1. X Liu1,
  2. H McGrath2,
  3. C Flanagan1 and
  4. L Zeng3
  1. 1University of Limerick, Limerick, Ireland
  2. 2University Hospital Limerick, Limerick, Ireland
  3. 3USETC, Chengdu, China


Background and Aims Artificial intelligence (AI) has been widely used in anaesthesiology, but recent advances promise to revolutionize its application in the field. Epileptic seizure prediction is clinically useful for patients with epilepsy, improving safety, increasing independence, and allowing for acute treatment.

Methods In this paper, eighteen AI algorithms were used in two different EEG datasets to predict epileptic seizures and obtained good results.

Results In the Bonn EEG database, ETC has the best test accuracy, SGDC has the smallest SD, and SVM has the highest F1 score; in the CHB-MIT Scalp EEG database, RF has the best test accuracy and the highest F1 score, SGDC has the smallest SD. The test accuracy of all artificial intelligence methods is above 75%, the standard deviation is less than 0.7, and the F1 score is above 0.06.

Conclusions The tree classifier may be the best predictor of epilepsy during anaesthesia in the EEG database. In the future, more AI algorithms suitable for epilepsy prediction will be further explored and verified. More unpopular but important AI algorithms will be applied to explore better ML solutions. AI could be a valuable ally for anaesthesiologists who want to increase their productivity and potentially improve their accuracy.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.