Waveform Extraction

Code for extracting waveform data from our server. Additionally, a basic PyTorch dataset class can be found in pytorch_dataset.py.

How to use

Data Extraction

This code can be run with the following command

python waveform_extraction.py --patient_list <your list of patients>

An example of the a patient list can be found in the repo named example_patient_list.csv.

PyTorch Dataset

First extract the data to CSVs using waveform_extraction.py

The dataset class uses a sliding window to grab frames from the waveform data.

The class has the following required arguments:

• root (str): path to data
• window_size (int): size of sliding window
• stride (int): stride size of sliding window

Additionally, this class allows you to pass in your own processing or PyTorch transform witht the following arguments:

• preprocess (function): preprocessing function
• transform (torch.transform): transform function

Lastly, data is split 80% training and 20% validation. The splits can be accessed using the following argument:

• train (bool):
  • if true, returns training split
  • if false, return test split

TODO

• Extract beats
• Preprocess