Towards an implantable closed-loop system for diagnosis and intervention in epilepsy

Epilepsy is a common neurological disease affecting about 1% of the world population, whereby about 30% of patients are resistant to antiepileptic drugs. As the minority of these patients are candidates for surgical treatment, effective seizure control remains challenging. A novel treatment method is the application of electrical stimulation during the early phases of a seizure, which serves to interrupt the spread of ictal activity across the brain. This can be achieved via a closed-loop system, whereby neuronal activity is recorded by a set of electrodes and continuously monitored by a seizure detection algorithm. When an emerging seizure is detected, electrical stimulation is triggered at the corresponding area. The present study therefore investigated how such a system can be realized. Firstly, four seizure detection algorithms were designed using the following methods: random forest (RF), support vector machine (SVM), multi-layer perceptron (MLP), and convolutional neural network (CNN). Evaluation of the iEEG recordings, showed that the proposed CNN model had the best performance, followed by the RF. The antiepileptic effects of electrical stimulation were then investigated via two approaches: 1. Responsive high-frequency (~250 Hz) stimulation of the seizure onset zone (SOZ) during the ictal phase via a closed-loop system 2. Interictal low-frequency (1 Hz) stimulation of the SOZ to gain a better understanding of the modulatory effects of cortical stimulation on excitability levels. This, in turn, may lead to improved stimulation protocols. The closed-loop system was found to be successful in terms of seizure detection, while the efficacy of stimulation varied between patients. The results revealed that low-frequency stimulation reduces excitability levels in the cortical tissue and may, in turn, exert antiepileptic effects. In conclusion, the results of this study have established a foundation for the development of a closed-loop system for implantable devices.