This directory contains a collection of Jupyter notebooks that demonstrate the data preparation and analysis workflow for the project.
get_dataset_statistics.ipynb- This notebook demonstrates how to get the prevalence of each clinical feature in the dataset. The prevalence of each type of myocardial scarring and the prevalence of LVEF<40% and LVEF<50% are also calculated.
xgb-with-clinical-features.ipynb- This notebook demonstrates how to train single-task XGBoost models on clinical features. The model is trained on clinical features of both old and new data. The resulting model can be used to predict either the presence of myocardial scarring or the LVEF range. This model is the baseline model.
evaluation.ipynb- This notebook demonstrates how to evaluate the performance of the models on the test set. The predicted probability from each model is generated here for further analysis.evaluation_1d.ipynb- This notebook demonstrates how to evaluate the performance of the models on the 1d test set. The predicted probability from each model is generated here for further analysis.
get_figures.ipynb- This notebook demonstrates how to generate the figures in the paper. This notebook uses the predicted probability generated from the evaluation notebook.get_figures_1d.ipynb- This notebook demonstrates how to generate the figures in the paper for 1d experiments. This notebook uses the predicted probability generated from the evaluation notebook.