Seizure Classifier

Dataset Information

To access the dataset we used, please visit: TUH EEG Corpus, and download version 1.4.0 of the dataset with rsync. This may require you to register and get an autogenerated username and password.

MATLAB Application Setup (For Testing/Debugging Purposes)

• Installing MATLAB Compiler (R2022a by MathWorks) for compiling the desktop application (ignore if already installed):
  • Select the "Home" tab upon opening MATLAB (towards the top left of the screen)
  • Click the arrow under the "Add-Ons" icon (towards the right of the screen) and click "Get Add-Ons"
  • In the Add-On explorer, search "MATLAB Compiler" and follow any instructions given to install it
• Select the "Apps" tab upon opening MATLAB (towards the top left of the screen)
• Select the "Design App" icon (located under the "Home" tab, also towards the top left of the screen)
• Click "Open..." and then navigate to the "SCC" folder to select "SCC.mlapp"
• Once "SCC.mlapp" is opened, the code as well as the design of the application can be modified
  • Select "Design View" (towards the top right of the application) to make any visual changes to the application
  • Select "Code View" to implement any additional functionalities and code
• Click "Run" to deploy and test out the application
• Once the changes are complete, be sure to click the "Save" icon
• Also be sure to click the arrow under the "Save" icon and select "Export to .m File..."
• Make sure there are actually changes made to "SCC.mlapp" and "SCC_exported.m" before pushing changes
• When ready to compile the desktop application:
  • Click the arrow under the "Share" icon and click "Standalone Desktop App"
  • Fill out the necessary information (e.g. application name, author names, summary, description, etc.)
  • When finished, click the "Package" icon (with the green check mark)

Neural Network

Once you have the dataset downloaded, everything should be able to run from the provided jupyter notebook after you designate a path for the preprocessed data to be stored. This will unfortunately require quite a bit of storage space (~20 gb).

SVM Documents

• Matlab R-2021b was used for the designing, training, and testing of this classifer.
• Initial toolboxes required are statistics and machine learning toolbox, signal processing toolbox, and the EEG feature extraction toolbox.
• The easiest way to run our code is to first download the entire SVM_final folder. Then save the path to the folder that is called SVM_functions which contains all the files for the important preprocessing and feature extraction prepared.
  • To actually run, train, and test the algorith, use the initial.m file, which is a script that will allow you to successfully train the SVM from the EDF files that are saved in the train folder. It will also test the SVM classifier using EDF files in the test folder. It is important you download the SVM foler the way it is and make sure to save a path to SVM_functions on your Matlab paths. The data used to train the SVM is from V 1.5.2 of the TUH dataset from the TUSZ corpus. Small sample data subset provided for training purposes, but rest of the data can be accessed from TUH database.
  • NOTE: It is very important to add the path to the SVM_functions folder as well as the toolboxes mentioned above, they contain files that are essential for proper functioning of the classifier.
  • Running it the way it is should allow recreation of results mentioned in the design document.