/docs/source/img/ECG-featurizer_banner.png

The purpose of this package is to make tabular data from ECG-recordings by calculating many features. The package is built on WFDB [1] and NeuroKit2 [2].

To install ECG-featurizer, run this command in your terminal:

pip install ECG-featurizer

from ECGfeaturizer import featurize as ef
# Make ECG-featurizer object
Feature_object =ef.get_features()
# Preprocess the data (filter, find peaks, etc.)
My_features=Feature_object.featurizer_dat(features=ecg_filenames,labels=labels,directory="./data/",demographical_data=demo_data)

from ECGfeaturizer import featurize as ef
number_of_ECGs = <the amount of ECGs>
directory = "<your dir>"
# Make ECG-featurizer object
Feature_object =ef.get_features()
# Preprocess the data (filter, find peaks, etc.)
My_features=Feature_object.featurizer_mat(num_features=number_of_ECGs, mat_dir = directory)

A numpy array of ECG-recordings in directory. Each recording should have a file with the recording as a time series and one file with meta data containing information about the patient and measurement information. This is standard format for WFDB and PhysioNet-files [1] [3]

Supported input files:

Input data Supported file format

ECG-recordings .dat files

Patient meta data .hea files

A numpy array of labels / diagnoses for each ECG-recording. The length of the labels-array should have the same length as the features-array .. code-block:: python

len(labels) == len(features)

A string with the path to the features. If the folder structure looks like this:

mypath

├── ECG-recordings

│ ├── A0001.hea

│ ├── A0001.dat

│ ├── A0002.hea

│ ├── A0002.dat

│ └── Axxxx.dat

then the feature and directory varaible could be:

features[0] "A0001"

directory "./mypath/ECG-recordings/"

The demographical data that is used in this function is age and gender. A Dataframe with the following 3 columns should be passed to the featurizer() function.

The strings in the filename_hr -column should be the same as the strings in the feature array. In this example gender is OneHot encoded such that

1 = Female 0 = Male

• A tutorial will come

• Some examples will come

GPLv3 license

Citation guidelines will come