Recent computer modeling techniques have made rendering photorealistic scenes possible in the context of gaming and animation. We sought to create a similar paradigm in the electroencephalogram (EEG) space in order to make accurate EEG interpretation more accessible to nonepileptologists. By combining data driven machine learning based approaches with better established physics models, we generated realistic synthetic EEG recordings. These synthetic recordings can be used in a variety of contexts including in creating denoising models, reconstructing missing channels, and augmenting existing datasets.

Code on git
Neural Fill
https://github.com/DanielLongo/NeuralFill
Data augmentation, denoising, physics based models
https://github.com/DanielLongo/eegML