FPGA Implementation of Epileptic Seizure Detection using Artificial Neural Network

Epilepsy is a neurological disorder accompanied by a sudden, uncontrolled electrical disturbance in the brain which is called seizures. So far, epilepsy has no actual treatment, but its symptoms and accompanying seizures are rather controlled by medication. This makes the continuous monitoring of epilepsy patients an important point to consider. Recently, machine learning and deep learning techniques are utilized in seizure prediction algorithms, which supports seizure monitoring significantly. Additionally, Field-programmable Gate Arrays (FPGAs) were used extensively in the past years to create hardware implementations of machine learning models to predict seizures. In this work, we aim to implement FPGA-based general and patient specific seizure prediction that detect seizures for all epilepsy patients using multilayer perceptron neural network models, which was not done before. This will be done by first extracting features from The raw electroencephalogram (EEG) signal samples provided by Melbourne-NeuroVista seizure trial and the Melbourne-University AES-MathWorks-NIH Seizure Prediction Challenge. Results show that the general model had an AUC score of 0.71 while the patient specific model had an AUC score of 0.79.