A Comparison of Machine Learning Classifiers for Energy-Efficient Implementation of Seizure Detection

Abstract

The closed-loop application of electrical stimulation via chronically implanted electrodes is a novel approach to stop seizures in patients with focal-onset epilepsy. To this end, an energy efficient seizure detector that can be implemented in an implantable device is of crucial importance. In this study, we first evaluated the performance of two machine learning algorithms (Random Forest classifier and support vector machine (SVM)) by using selected time and frequency domain features with a limited need of computational resources. Performance of the algorithms was further compared to a detection strategy implemented in an existing closed loop neurostimulation device for the treatment of epilepsy. The results show a superior performance of the Random Forest classifier compared to the SVM classifier and the reference approach. Next, we implemented the feature extraction and classification process of the Random Forest classifier on a microcontroller to evaluate the energy efficiency of this seizure detector. In conclusion, the feature set in combination with Random Forest classifier is an energy efficient hardware implementation that shows an improvement of detection sensitivity and specificity compared to the presently available closed-loop intervention in epilepsy while preserving a low detection delay.

Keywords: closed-loop intervention; epilepsy; intracranial EEG; low power microcontroller implementation; machine learning; seizure detection.

Figure 1

Detection performance of the classifiers over all the patients based on a best channel (highest area under the receiver operating characteristic (ROC) curve (AUC)) selection per patient.

Figure 2

Early detection performance of the classifiers over all the patients based on a best channel (highest AUC) selection per patient. Here, just the seizures which could be detected in the first 10 s are counted as a true positive.

Figure 3

Detection performance of the classifiers over all the patients in multichannel mode.

Figure 4

Early detection performance of the classifiers over all the patients in multichannel mode. Here, just the seizures which could be detected in the first 10 s are counted as a true positive.

Figure 5

ROC curves of all patients in single and multichannel mode for the Random Forest classifier. The dashed line separates the early retrieval area.

Figure 6

ROC curves of all patients in single and multichannel mode for the support vector machine (SVM) classifier. The dashed line separates the early retrieval area.

Figure 7

ROC curves of all patients in single and multichannel mode for the line length classifier. The dashed line separates the early retrieval area.

Figure 8

Required runtime in number of cycles for each feature.