Fill Database WaveForm Headers

1) Import de las librerias que utilizaremos



In [1]:

    
import urllib.request
import wfdb
import psycopg2
from psycopg2.extensions import AsIs

2) Leemos el archivo con las WaveForm que vamos a utilizar



In [2]:

    
target_url = "https://physionet.org/physiobank/database/mimic3wdb/matched/RECORDS-waveforms"
data = urllib.request.urlopen(target_url) # it's a file like object and works just like a file
lines = data.readlines();
line = str(lines[2])
line









    Out[2]:





"b'p00/p000033/p000033-2116-12-24-12-35\\n'"

3) Limpiamos los caracteres extraños y Dividimos la cadena donde pXXNNNN-YYYY-MM-DD-hh-mm donde XXNNNN es el identificador unico del paciente SUBJECT_ID y YYYY-MM-DD-hh-mm es la fecha de la estadia del paciente'



In [3]:

    
line = line.replace('b\'','').replace('\'','').replace('\\n','')
splited = line.split("/")
carpeta,subCarpeta,onda = line.split("/")
carpeta = carpeta+"/"+subCarpeta
subject_id = subCarpeta.replace('p','')
recordDate = onda.replace(subCarpeta+"-","")
print("subject_id: ",subject_id)
print("recordDate: ",recordDate)
print("onda: ",onda)
print("carpeta: ",carpeta)









    



subject_id:  000033
recordDate:  2116-12-24-12-35
onda:  p000033-2116-12-24-12-35
carpeta:  p00/p000033

4) Leemos el encabezado del waveform, para obtener la información del paciente que almacenaremos



In [4]:

    
try:
    sig, fields = wfdb.srdsamp(onda,pbdir='mimic3wdb/matched/'+carpeta, sampto=1)
    print(fields)
except Exception as inst:
    print("onda vacia")









    



{'units': ['mV', 'mV'], 'signame': ['II', 'MCL1'], 'comments': [], 'fs': 125}

Le agregamos el subject_id y la fecha del record a los campos



In [5]:

    
fields['subject_id'] = subject_id
fields['recordDate'] = recordDate

convertimos los campos en un diccionario



In [6]:

    
columns = fields.keys()
values = [fields[column] for column in columns]
print(columns)









    



dict_keys(['units', 'signame', 'comments', 'recordDate', 'subject_id', 'fs'])

Nos conectamos a la base de datos postgres donde almacenaremos los datos



In [7]:

    
conn = psycopg2.connect("dbname=mimic")
cur = conn.cursor()

Creamos la tabla donde quedaran almacenados los datos



In [8]:

    
table = "waveformFields"
#cur.execute("DROP TABLE "+table)
cur.execute("CREATE TABLE IF NOT EXISTS "+table+
            " (id serial PRIMARY KEY,"+
            "comments character varying(255)[],"+
            "fs integer, signame character varying(255)[],"+
            "units character varying(255)[],"+
 "subject_id integer,"+
            "recordDate character varying(255));")

verificamos si ya existe el dato



In [9]:

    
def track_not_exists(cur, subject_id,recordDate):
    select_stament = 'select id from '+table+' where subject_id= %s and recorddate = %s'
    cur.execute(select_stament,(int(subject_id),recordDate))
    return cur.fetchone() is None
#print(cur.mogrify(select_stament,(int(subject_id),recordDate)))
notExist = False
if track_not_exists(cur,subject_id,recordDate):
    notExist = True
    print("not exist")

Insertamos los datos



In [10]:

    
insert_statement = 'insert into '+table+' (%s) values %s'

print(cur.mogrify(insert_statement, (AsIs(','.join(columns)), tuple(values))))
if notExist:
    cur.execute(insert_statement, (AsIs(','.join(columns)), tuple(values)))









    



b"insert into waveformFields (units,signame,comments,recordDate,subject_id,fs) values (ARRAY['mV', 'mV'], ARRAY['II', 'MCL1'], '{}', '2116-12-24-12-35', '000033', 125)"

Hacemos commit



In [11]:

    
conn.commit()

cerramos conexion



In [12]:

    
conn.close()