Seizures (Epileptic Seizures in specific) are the second most common neurological disorder and affect tens of millions of people each year. Seizure detection is generally a highly involved process requiring medical experts on hand. Automatic seizure detection has developed into an active area of research, but faces a number of difficulties. EEG (electroencephelogram) recordings are how brain activity is generally recorded for seizure detection. Automatic seizure detection generally relies on hand-crafted feature extraction from EEG data - a difficult and time-intensive process. Applying Machine learning techniques such as Convolutional Neural Networks could potentially do away with hand-crafted features, in favor of learned features.

In order to pass our EEG data to a CNN, we first need to convert it into a spectrogram, which gives us a visual representation of frequency over time. Our network architecture will look like:

Code implementation based on following papers:

1. Truong, N., Nguyen, A., Kuhlmann, L., Bonyadi, M., Yang, J., Ippolito, S. and Kavehei, O. (2018). Convolutional neural networks for seizure prediction using intracranial and scalp electroencephalogram. Neural Networks, 105, pp.104-111. and their code implementation: https://github.com/NeuroSyd/seizure-prediction-CNN/

2. Zhou, M., Tian, C., Cao, R., Wang, B., Niu, Y., Hu, T., … Xiang, J. (2018). Epileptic Seizure Detection Based on EEG Signals and CNN. Frontiers in neuroinformatics, 12, 95. doi:10.3389/fninf.2018.00095