Summer project - University of Toronto, IMS SURP
Abigail Wolfensohn, Timur Latypov, Hodaie Lab
Primary objective: Using Freesurfer parcellation to predict TN pain in MS subjects
- data_preprocessing.py - retrieves and reorganizes raw MS and MS-TN demographic and featural data so that it can be used by the machine learning model
- tSNE.py - performs dimensionality reduction and visualizes the data structure
- model.py - runs SVM with sequential feature selection to predict TN pain in MS; nested k-fold cross-validation
- graphs.ipynb - runs graphic representations of data gathered by the model, as well as an independent t-test
- stats - folder for csv files containing featural (from Freesurfer segmentation) and demographic data for each subject
- utils - folder for other important files to be used by the model and graphing notebook
- out - folder for output files
Mean train accuracy 99.5%
Mean test accuracy 93.4%
Manuscript is in preparation.