An Energy Efficient Seizure Prediction Algorithm
Epileptic seizures afflict over 1% of the world’s population. If seizures could be predicted before they occur, fast-acting therapies could be delivered to prevent the attack and restore a normal quality of life to patients. Over the last two decades, several studies have explored the use of EEG signals to predict seizures using principles from machine learning [1]–[3]. It is thought that such an algorithm could be implemented in real-time with a wireless, implanted EEG sensor. However, there are two main constraints for such a portable system. First, due to limited battery life, energy consumption must be minimal. Second, due to limited bandwidth, the data transmitted between the sensor and the central processing device (such as mobile phone, tablet, personal computer, etc.) should be small. To address these issues, we sought to develop a robust learning algorithm that identifies EEG time series as pre-ictal (the time just before a seizure occurs) or interictal (the time between seizures) using downsampled data. This could ultimately reduce both power consumption and bandwidth usage for wearable seizure prediction devices
Research Paper Link: Download Paper