Help

Epi-AI allows the user to choose the EEG file that contains a single-channel EEG signal. Epi-AI will predict if the events are seizure events or non-seizure events. Epi-AI then returns the start time and end time of the detected seizure events. Meanwhile, the amplitude and corresponding spectrogram from the start time to the end time (in seconds) of each detected seizure event can be downloaded for the user to analyse further.

We have chosen one of the Intra-amygdala kainic acid models of epilepsy (C57BL/6) as an example. This example EEG data was not included in our training data and can be downloaded from "Sample EEG files" on the Epi-AI webserver (Model_III.edf). First, this EDF file contains two-channel EEG recordings, please choose one of the channel you want to analyse (channel 1 or 2), and then upload the Model_III.edf and click the "Seizure detection" button. Next, the start, end time and the duration of each detected seizure event can be downloaded in the new page (by clicking 'Download the result'). The amplitude and corresponding spectrogram from the start time to the end time (in seconds) of each detected seizure events can also be downloaded by clicking the "Download the result" for the user to further analyse. This example data contains 1 hour of EEG recordings, which would take time for experts to annotate. However, Epi-AI takes around 20 s to process the single-channel EEG file and then return the results to the user. We also provide some other mouse models (Model I - IV) which we used to develop the Epi-AI, these files can be downloaded from "Sample EEG files".

Only three mouse models (Intra-amygdala kainic acid adult male SV129 mice model of epilepsy; Dravet syndrome model of epilepsy; Intra-amygdala kainic acid adult male C57BL/6 mice model of epilepsy;) were used to train Epi-AI, the pilocarpine model was not used in training. The Epi-AI webserver has been tested with four independent test sets, including a pilocarpine independent test set. It is possible that Epi-AI will also perform well with EEG data from other mouse models of epilepsy, however, we have not tested this. If user would like to convert the EEG file from other formats to EDF/PICKLE/CSV files, or they would like to know the sampling frequency of their data, they can visit Github for help. If a user would like to use Epi-AI with data from another mouse model then either the user can test this themselves first, or they can contact us (Catherine.Mooney@ucd.ie) for assistance. In addition, we will make the executables available for academic use if any labs would like to contact us. We will assist researchers in installing a local version of Epi-AI.