search

neurosimata / seizy_ml

Automated seizure detection coupled with manual validation
Apache License 2.0
0 stars 0 forks source link

Example files #1

Open NathanTBurgess opened 1 month ago

NathanTBurgess commented 1 month ago

Hello! I was wondering if you could provide example data.

Based on what I'm reading in the training tutorial all training data should be concatenated into one .h5 file along with an label file. What is the format of this label file? will each time point be labeled as seizure or non-seizure?

The workflow, as I am currently thinking of it.

  1. convert a set of edf files to on h5 file

  2. create an h5 file that matches the length of the concatenated h5 file made from edf files which identifies (at each time point) whether a seizure is occurring or not.

  3. follow the training tutorial (the preprocessing seems to be built into the train function)

  4. follow the main tutorial

pantelisantonoudiou commented 3 weeks ago

Hi thank you for you question and sorry for the late reply! I did not receive a notification for this issue, but I have changed the settings now so I will be receiving new notifications.

Did you manage to get the workflow working? All the steps you highlighted are correct. One more thing that I found makes a big difference is normalizing the extracted features per file before concatenating.

In the next couple of weeks or so I will be 1) updating seizyML to require each h5 recording file separately for training (plus adding a few more app features) and 2) I will create a simple repo for converting edf files to h5.