MIT-BIH心率失常数据的读取以及简单处理

导入第三方包/设置路径/定义和数据选择

# -*- coding: utf-8 -*-
import wfdb
import numpy as np
import matplotlib.pyplot as plt
from scipy.signal import medfilt
import pywt
import pandas as pd

path_load1 = 'F:/ECG_MIT-BIH/MIT-BIH/'
path_training_save1 = 'F:/ECG_MIT-BIH/training/path_save1/'
path_load2 = 'F:/ECG_MIT-BIH/training/path_save1/'
path_training_save2 = 'F:/ECG_MIT-BIH/training/path_save2/'
path_training_save3 = 'F:/ECG_MIT-BIH/training/path_save3/'
path_load3 = 'F:/ECG_MIT-BIH/training/path_save3/'
path_training_save4 = 'F:/ECG_MIT-BIH/training/path_save4/'
path_load4 = 'F:/ECG_MIT-BIH/training/path_save4/'
path_training_lead0_save5 = 'F:/ECG_MIT-BIH/training/path_save5_double/lead0/'
path_training_lead1_save5 = 'F:/ECG_MIT-BIH/training/path_save5_double/lead1/'

DS1 = ['101', '106', '108', '109', '112', '114', '115', '116', '118','119', '122',
       '124', '201', '203', '205', '207', '208', '209', '215', '220', '223', '230']
DS2 = ['100', '103', '105', '111', '113', '117', '121', '123', '200', '202', '210', '212', 
       '213', '214', '219', '221', '222', '228', '231', '232', '233', '234']
DS = DS1          #DS1, DS2
DSName = 'DS1'    #DS1, DS2

原始数据转存为.csv文件

# 原始数据的读取
def ReadTrainingSet(DS):
    for ds in DS:
        dataSetLoc = path_load1 + ds
        record = wfdb.rdrecord(dataSetLoc)
        lead0 = record.p_signal[:,0]
        lead1 = record.p_signal[:,1]
        annotation = wfdb.rdann(dataSetLoc, 'atr')
        annIdx = annotation.sample
        annType = annotation.symbol
        trainingSet[ds] = {'lead0': lead0, 'lead1': lead1, 'annIdx': annIdx, 'annType': annType}

trainingSet = {} 
ReadTrainingSet(DS1)

# 将数据保存为.csv文件，位置为 path_training_save1
headers = ['lead0', 'lead1', 'annIdx', 'annType']

for k, v in trainingSet.items():
    df1 = pd.DataFrame({ 'lead0': trainingSet[k]['lead0'], 'lead1': trainingSet[k]['lead1']})
    df1.to_csv( path_training_save1 + k + '_lead.csv', index=False)
    
    df2 = pd.DataFrame({ 'annIdx': trainingSet[k]['annIdx'], 'annType': trainingSet[k]['annType']})
    df2.to_csv( path_training_save1 + k + '_ann.csv', index=False)
    print(k,end = '')

# 检测数据的读取情况
plt.figure(figsize=(20, 6), dpi=80)
temp1 = trainingSet['109']['lead0'][:6000]
plt.plot(temp1)
plt.show()

结果如下

心拍分割

# 读取刚刚保存的数据
trainingSet_leads = {}
trainingSet_anns = {}
for ds in DS:
    trainingSet_leads[ds] = pd.read_csv(path_load2 + ds + '_lead.csv')
    trainingSet_anns[ds] = pd.read_csv(path_load2 + ds + '_ann.csv')
    print(ds,end = '')
print('\n读取完毕！')

# 心拍分割
Hbs = {}
for ds in DS:
    lead0 = trainingSet_leads[ds]['lead0'] #取lead0列
    lead1 = trainingSet_leads[ds]['lead1'] #取lead1列
    hbs0 = []
    hbs1 = []
    anns = []
    annIdxs = []
    for row in trainingSet_anns[ds].itertuples():
        if row[2] in Non_beat_anns:
            continue #跳出这轮循环，这个数据不要
        elif row[1] < 91:
            continue #跳出这轮循环，说明是第一个时间点
        elif row[1] + 144 > len(lead0):
            continue #
        else:
            anns.append(row[2])
            annIdxs.append(row[1] - 1)
            hbs0.append(lead0[row[1] - 91: row[1] + 144]) #加入1个RR间期
            hbs1.append(lead1[row[1] - 91: row[1] + 144]) #加入1个RR间期
    Hbs[ds] = pd.DataFrame({'lead0': hbs0, 'lead1': hbs1, 'ann': anns, 'annIdx': annIdxs})
    print(ds,end='')
print('\n分割完成！')

# DS1记录114导联校正：
if DSName == 'DS1':
    temp = pd.Series(list(Hbs['114']['lead0']))
    Hbs['114']['lead0'] = Hbs['114']['lead1']
    Hbs['114']['lead1'] = temp
print('校正完成！')

# 分成lead0和lead1两个数据集
Hbs_lead0 = {}
Hbs_lead1 = {}

for ds in DS:
    Hbs_lead0[ds] = pd.DataFrame({
        'ann': Hbs[ds]['ann'],
        'annIdx': Hbs[ds]['annIdx'],
        'beat': Hbs[ds]['lead0'],
    })
    
    Hbs_lead1[ds] = pd.DataFrame({
        'ann': Hbs[ds]['ann'],
        'annIdx': Hbs[ds]['annIdx'],
        'beat': Hbs[ds]['lead1'],
    })
print('完成！')

# 将以上划分的数据保存为CSV文件，位置为 path_training_save2
for ds in DS:
    beatValues = []
    beatIndexs = []
    for i in range(len(Hbs_lead0[ds]['beat'])):
        beatValues.append(list(Hbs_lead0[ds]['beat'][i].values))
        beatIndexs.append(list(Hbs_lead0[ds]['beat'][i].index))
    Hbs_lead0[ds]['beatValues'] = beatValues
    Hbs_lead0[ds]['beatIndex'] = beatIndexs
    
    beatValues = []
    beatIndexs = []
    for i in range(len(Hbs_lead1[ds]['beat'])):
        beatValues.append(list(Hbs_lead1[ds]['beat'][i].values))
        beatIndexs.append(list(Hbs_lead1[ds]['beat'][i].index))
    Hbs_lead1[ds]['beatValues'] = beatValues
    Hbs_lead1[ds]['beatIndex'] = beatIndexs
    Hbs_lead0[ds].to_csv( path_training_save2 + ds + '_lead0.csv', index=False)
    Hbs_lead1[ds].to_csv( path_training_save2 + ds + '_lead1.csv', index=False)
    print(ds,end = '')
print('\n已保存到path_training_save2中！')

# 整理数据（幅值+标签）并保存，位置为 path_training_save3
Hbs_lead0_0 = {}
Hbs_lead1_1 = {}
for ds in DS:
    Hbs_lead0_0[ds]=Hbs_lead0[ds].drop(columns = ['annIdx','beat','beatIndex'])
    Hbs_lead1_1[ds]=Hbs_lead1[ds].drop(columns = ['annIdx','beat','beatIndex'])
    Hbs_lead0_0[ds].to_csv( path_training_save3 + ds + '_lead0.csv', header = 0, index=False) #不保留列名，索引
    Hbs_lead1_1[ds].to_csv( path_training_save3 + ds + '_lead1.csv', header = 0, index=False) #不保留列名，索引
print('保存完毕！')

# 数据分列并保存，位置为 path_training_save4
Hbs_lead0 = {}
Hbs_lead1 = {}

for ds in DS:
    Hbs_lead0[ds] = pd.read_csv( path_load3 + ds + '_lead0.csv',header = None,names = ['A','B'])
    Hbs_lead1[ds] = pd.read_csv( path_load3 + ds + '_lead1.csv',header = None,names = ['A','B'])

for key, value in Hbs_lead0.items():
    df0 = pd.DataFrame({ 'index': Hbs_lead0[key]['A'], 'lead0': Hbs_lead0[key]['B']})
    df1 = df0['index'].str.split(',',expand=True)
    df2 = df0['lead0'].str.lstrip("[")
    df2 = df2.str.rstrip("]")
    df2 = df2.str.split(',',expand=True)
    df = pd.concat([df1,df2],join="inner",axis=1)  
    df.to_csv( path_training_save4 + key + '_lead0.csv',header=0,index=None)
    print(key,end = '')
print('\nlead0保存完毕！')

for key, value in Hbs_lead1.items():
    df0 = pd.DataFrame({ 'index': Hbs_lead1[key]['A'], 'lead1': Hbs_lead1[key]['B']})
    df1 = df0['index'].str.split(',',expand=True)
    df2 = df0['lead1'].str.lstrip("[")
    df2 = df2.str.rstrip("]")
    df2 = df2.str.split(',',expand=True)
    df = pd.concat([df1,df2],join="inner",axis=1)  
    df.to_csv( path_training_save4 + key + '_lead1.csv',header=0,index=None)
    print(key,end = '')
print('\nlead1保存完毕！')

重新标注（可以2分类或5分类）

# 心跳注释
N = ['N', 'L', 'R', 'e', 'j']
S = ['A', 'a', 'J', 'S']        #SVEB
V = ['V', 'E']                  #VEB
F = ['F']
Q = ['l', 'f', 'Q']
Non_beat_anns = ['[', ']', '!', 'x', '(', ')', 'p', 't', 'u', '`', '~', '^', '|', '+', 's', 'T', '*', 'D', '=', '"', '@']

# 重新标注
Hbs_lead0 = {}
Hbs_lead1 = {}

for ds in DS:
    Hbs_lead0[ds] = pd.read_csv( path_load4 + ds + '_lead0.csv',header = None)
    Hbs_lead1[ds] = pd.read_csv( path_load4 + ds + '_lead1.csv',header = None)
for key, value in Hbs_lead0.items():
    df0 = pd.DataFrame({ 'index': Hbs_lead0[key].loc[:,0]})
# df0
# type(df0)
    for i in range(0,len(df0['index'])):
        if df0['index'][i] in N:
            df0['index'][i] = 0
        elif df0['index'][i] in S:
            df0['index'][i] = 1
        elif df0['index'][i] in V:
            df0['index'][i] = 1
        elif df0['index'][i] in F:
            df0['index'][i] = 1
        elif df0['index'][i] in Q:
            df0['index'][i] = 1      
    df1 = Hbs_lead0[key][list(range(236))]
    df1 = df1.drop(columns=0,axis=1)
    df = pd.concat([df0,df1],join="inner",axis=1)
    df.to_csv( path_training_lead0_save5 + key + '_lead0.csv',header=0,index=None)
    print(key,end = '')
print('\nlead0保存完毕！')

for key, value in Hbs_lead1.items():
    df0 = pd.DataFrame({ 'index': Hbs_lead1[key].loc[:,0]})
    for i in range(0,len(df0['index'])):
        if df0['index'][i] in N :
            df0['index'][i] = 0
        elif df0['index'][i] in S:
            df0['index'][i] = 1
        elif df0['index'][i] in V:
            df0['index'][i] = 1
        elif df0['index'][i]in F:
            df0['index'][i] = 1
        elif df0['index'][i] in Q:
            df0['index'][i] = 1
    df1 = Hbs_lead1[key][list(range(236))]
    df1 = df1.drop(columns=0,axis=1)
    df = pd.concat([df0,df1],join="inner",axis=1)
    df.to_csv( path_training_lead1_save5 + key + '_lead1.csv',header=0,index=None)
    print(key,end = '')
print('\nlead1保存完毕！')