Bsingstad/ECG-featurizer

A python package for extracting features from ECGs

ECG-featurizer

A method to extract features from electrocardiographic recordings

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].

Installation

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

pip install ECG-featurizer

Documentation:

Featurize .dat-files:

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)

Featurize .mat-files:

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)

features:

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

labels:

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)

directory:

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/"

demographical_data:

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.

	age	gender	filename_hr
0	11.0	1	"A0001"
1	57.0	0	"A0002"
2	94.0	0	"A0003"
3	34.0	1	"A0004"

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

Tutorials:

• A tutorial will come

Other examples:

• Some examples will come

Contributing

Citation:

Citation guidelines will come

Popularity:

References:

[1]	WFDB: https://github.com/MIT-LCP/wfdb-python

[2]	Makowski, D., Pham, T., Lau, Z. J., Brammer, J. C., Lesspinasse, F., Pham, H., Schölzel, C., & S H Chen, A. (2020). NeuroKit2: A Python Toolbox for Neurophysiological Signal Processing. Retrieved March 28, 2020, from https://github.com/neuropsychology/NeuroKit

[3]

Goldberger AL, Amaral LAN, Glass L, Hausdorff JM, Ivanov PCh, Mark RG, Mietus JE, Moody GB, Peng CK, Stanley HE. PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals. Circulation 101(23):e215-e220 [Circulation Electronic Pages; http://circ.ahajournals.org/content/101/23/e215.full]; 2000 (June 13). PMID: 10851218; doi: 10.1161/01.CIR.101.23.e215