Experiment code for Efficient Self-supervised Pretraining with Simple Synthesized ECG

Prepare directories

• PATH_TO_ORIGINAL_DATA: Edit line7 of invoke_container.sh
• PATH_TO_PROCESSED_DATA_SAVE_DIR: Edit line8 of invoke_container.sh

Data preparation

1. Real-world data.

• Download CPSC, G12EC, and PTBXL dataset and place at PATH_TO_ORIGINAL_DATA
• Move to src/prep/dataset and run bash prep_data.sh.

2. Synthesized data.

• Move to src/prep/syn and run bash syn_data.sh

Baseline selection

• Move to src/baselines
• cd resources and run python generate_yaml.py 1-19
• cd .. and run python experiment.py --exp 1-19

MAE pretraining

• Move to src/mae_pt
• cd resources and run python prepare_pt_yamls.py 1p
• cd .. and run python pretrain.py --pt <pretrain_id>

Abnormal ECG classification

• Move to src/mae_exp
• cd resources and run python generate_yaml.py
• cd .. and run python experiment.py --exp <exp_id>