physionet_challenge_utility_script.py

import os
import numpy as np, sys,os
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from scipy.io import loadmat
import wfdb
import tarfile
from sklearn.preprocessing import MultiLabelBinarizer
import math

import matplotlib.pyplot as plt
import seaborn as sns


def load_challenge_data(filename):
    x = loadmat(filename)
    data = np.asarray(x['val'], dtype=np.float64)
    new_file = filename.replace('.mat','.hea')
    input_header_file = os.path.join(new_file)
    with open(input_header_file,'r') as f:
        header_data=f.readlines()
    return data, header_data

def import_key_data_Georgia(path):
    gender=[]
    age=[]
    labels=[]
    ecg_filenames=[]
    for subdir, dirs, files in sorted(os.walk(path)):
        for filename in files:
            filepath = subdir + os.sep + filename
            if filepath.endswith(".mat"):
                data, header_data = load_challenge_data(filepath)
                if data.shape[1] == 5000:
                    labels.append(header_data[15][5:-1])
                    ecg_filenames.append(filepath)
                    gender.append(header_data[14][6:-1])
                    age.append(header_data[13][6:-1])
    return gender, age, labels, ecg_filenames

def import_key_data_China(path):
    gender=[]
    age=[]
    labels=[]
    ecg_filenames=[]
    for subdir, dirs, files in sorted(os.walk(path)):
        for filename in files:
            filepath = subdir + os.sep + filename
            if filepath.endswith(".mat"):
                data, header_data = load_challenge_data(filepath)
                if data.shape[1] > 5000:
                    labels.append(header_data[15][5:-1])
                    ecg_filenames.append(filepath)
                    gender.append(header_data[14][6:-1])
                    age.append(header_data[13][6:-1])
    return gender, age, labels, ecg_filenames

def make_undefined_class(labels, df_unscored):
    df_labels = pd.DataFrame(labels)
    for i in range(len(df_unscored.iloc[0:,1])):
        df_labels.replace(to_replace=str(df_unscored.iloc[i,1]), inplace=True ,value="undefined class", regex=True)

    '''
    #equivalent classes
    codes_to_replace=['713427006','284470004','427172004']
    replace_with = ['59118001','63593006','17338001']

    for i in range(len(codes_to_replace)):
        df_labels.replace(to_replace=codes_to_replace[i], inplace=True ,value=replace_with[i], regex=True)
    '''
    return df_labels

def onehot_encode(df_labels):
    one_hot = MultiLabelBinarizer()
    y=one_hot.fit_transform(df_labels[0].str.split(pat=','))
    y = np.delete(y, -1, axis=1)
    return y, one_hot.classes_[0:-1]