BEAT-PD Challenge
Training of individual classifiers for medication status, dyskinesia and tremor.
Required folder structure:
- The scripts will load the data from a folder structure as follows:
- "..\Datasets$study$dataset", where $study can take the values 'CIS' and 'REAL', and $dataset can take the values "training_data" and "testing_data", as provided in the ".tar" files.
- The ".csv" files containing the files and subjects IDs vinculation should be in the same folder as the segmented data files.
prepare_time_series.py:
- This script will read the segmented data files, calculate the features, and store the processed data in an ".hdf5" file
- The arguments needed are "--study" and "--dataset". "--study" can take the values "CIS" or "REAL", and "--dataset" can take the values "Train" or "Test";
- The output hdf5 file is named "$study_$dataset.hdf5", where $study is the argument passed as "--study", and $dataset is the argument passed as "--dataset".
train_predict.py:
- This script will train the model based on the "$study_Train.hdf5" files, generated by the prepare_time_series.py script, predict the labels for the observations stored in the "$study_Test.hdf5" files, and output a ".csv" file with the predictions;
- The arguments needed are "--study" and "--symptom". "--study" can take the values "CIS" or "REAL", and "--symptom" can take the values "medication", "dyskinesia" or "tremor";
- The output csv file is named "$study_predicts.csv" and will contain a table with the columns measurement_id, subject_id, and predicted_label