Code for the master thesis "Topological Data Analysis on Multimodal Brain Data" by Pia Baronetzky (supervisors Bastian Rieck and Silviu Bodea).
This code is tested on a Mac with Python version 3.10.12.
For installing the packages and dependencies, run
poetry install
The Anesthesia_Data folder contains code for classifying multimodal brain data into anesthesia stages.
Within the Anesthesia_Data folder, the Time_Series folder contains code for creating machine learning features from the EEG/EMG part of the data. The Brain_Imaging folder contains code for creating features from the brain imaging part of the data.
The Sleep_Data folder contains code for classifying EEG/EMG data into sleep stages.
Utils contains helpers for both data types.
Both the Anesthesia_Data and Sleep_Datafolders require a folder Data containing the data, as well as folders Embeddings_and_Persistence_Diagrams, Features, Plots and Train_Test_Splitting. Some of these folders require subfolders for the single subjects, but this should become clear from the code.
In both folders Anesthesia_Data and Sleep_Data, there is a file Classification.ipynbfor the main classification (and additionally, a fileClassification_Statistical_Features.ipynb` for the baseline classification methods using statistical features). Thes classification files depend on features that first have to be generated.
In order to generate the features, run the files Preprocessing_And_Computing_Persistence_Diagrams.ipynb first. Then, run all files ending with _Features.ipynb. You can then finally run the Classification.ipynb files.
The Data_Exploration.ipynb require the Compute_Wasserstein_Barycenters.ipynb to be run first.