MIMIC-IV数据分析 - 使用Python进行心脏病案例分析（二）

阅读本文请先看 MIMIC-IV数据分析 - 使用Python进行心脏病案例分析

五、图 4：患者的意识水平是多少？

• 格拉斯哥昏迷量表 (GCS) 是一种意识测量方法。(神经功能的衡量标准)

  • 范围从 3（最差，昏迷）到 15（最好，正常功能）。

• 它通常用于监测重症监护病房的患者

• 它由三个部分组成：眼睛反应；口头回应；运动反应。

import numpy as npimport pandas as pdimport matplotlib.pyplot as pltimport psycopg2
schema_name = 'mimic'# 连接到MIMIC-IV数据库conn = psycopg2.connect(dbname='mimiciv', user='postgres', password='mimic',                        host='10.241.148.228', port=5432)query_schema = 'SET search_path to ' + schema_name + ';'
# # # 设置查询语句# # # 我们选择从mimiciv_hosp.admissions表中提取hadm_id等于10006的行。# # # 在写sql代码时，最好先执行“set search_path to mimiciv" 随后的所有操作均不需要指明表格的位置；否则，任何操作都应该在表格名前面加前缀mimiciv# query1 = query_schema + 'SELECT subject_id, hadm_id, admittime, dischtime, admission_type FROM  mimiciv_hosp.admissions'## # 运行查询并将结果分配给变量# admissions_pd = pd.read_sql_query(query1,conn)# admissions_pd.head()# print(admissions_pd.head())
query = """

SELECT de.stay_id  , mimiciv_derived.datetime_diff(de.charttime, ie.intime, 'HOUR' ) AS HOURS  , di.label  , de.value  , de.valuenum  , de.valueuomFROM mimiciv_icu.chartevents deINNER join  mimiciv_icu.d_items diON de.itemid = di.itemidINNER join  mimiciv_icu.icustays ieON de.stay_id = ie.stay_idWHERE de.stay_id = 30121190ORDER BY charttime;

"""
ce = pd.read_sql_query(query, conn)
# OPTION 2: load chartevents from a CSV file# ce = pd.read_csv('data/example_chartevents.csv', index_col='HOURSSINCEADMISSION')ce.head()
# ce.label# print(ce.label)

# Select just the heart rate rows using an index

d = ce[ce['label'].str.contains('Heart Rate')]# Display the first few rows of the GCS eye response datagcs = ce[ce['label'].str.contains('GCS - Eye Opening')]gcs_m = ce[ce['label'].str.contains('GCS - Motor Response')]gcs_v = ce[ce['label'].str.contains('GCS - Verbal Response')]print(gcs)
r = ce[ce['label'].str.contains('Respiratory Rate')]

# Prepare the size of the figureplt.figure(figsize=(12, 10))
# # Set x equal to the timesx_hr = d['hours']## # Set y equal to the heart ratesy_hr = d['valuenum']## # Plot time against heart rateplt.plot(x_hr, y_hr)
plt.plot(r['hours'],         r['valuenum'],         'k', markersize=6)
# Add a text label to the y-axisplt.text(-20,155,'GCS - Eye Opening',fontsize=14)plt.text(-20,150,'GCS - Motor Response',fontsize=14)plt.text(-20,145,'GCS - Verbal Response',fontsize=14)
# Iterate over list of GCS labels, plotting around 1 in 10 to avoid overlapfor i, txt in enumerate(gcs['value'].values):    if np.mod(i,6)==0 and i < 65:        plt.annotate(txt, (gcs['hours'].values[i],155),fontsize=14)
for i, txt in enumerate(gcs_m['value'].values):    if np.mod(i,6)==0 and i < 65:        plt.annotate(txt, (gcs_m['hours'].values[i],150),fontsize=14)
for i, txt in enumerate(gcs_v['value'].values):    if np.mod(i,6)==0 and i < 65:        plt.annotate(txt, (gcs_v['hours'].values[i],145),fontsize=14)
plt.title('Vital signs and Glasgow Coma Scale over time from admission',fontsize=16)
plt.xlabel('Time (hours)',fontsize=16)plt.ylabel('Heart rate or GCS',fontsize=16)plt.ylim(10,165)plt.show()

患者的意识随着时间的推移如何变化？

我们换个患者看看。WHERE de.stay_id = 36648401

关注【科研收录】回复"患者意识水平"获取以上代码

六、图 5：患者输出体液随时间变化情况

我们现在将查询 outputevents 输出事件表 该表包含有关患者输出的所有信息（尿液输出、引流、透析）。

import numpy as npimport pandas as pdimport matplotlib.pyplot as pltimport psycopg2
schema_name = 'mimic'# 连接到MIMIC-IV数据库conn = psycopg2.connect(dbname='mimiciv', user='postgres', password='mimic',                        host='10.241.148.228', port=5432)query_schema = 'SET search_path to ' + schema_name + ';'
# # # 设置查询语句# # # 我们选择从mimiciv_hosp.admissions表中提取hadm_id等于10006的行。# # # 在写sql代码时，最好先执行“set search_path to mimiciv" 随后的所有操作均不需要指明表格的位置；否则，任何操作都应该在表格名前面加前缀mimiciv# query1 = query_schema + 'SELECT subject_id, hadm_id, admittime, dischtime, admission_type FROM  mimiciv_hosp.admissions'## # 运行查询并将结果分配给变量# admissions_pd = pd.read_sql_query(query1,conn)# admissions_pd.head()# print(admissions_pd.head())
query = """

SELECT de.stay_id  , mimiciv_derived.datetime_diff(de.charttime, ie.intime, 'HOUR' ) AS HOURS  , di.label  , de.value  , de.valueuomFROM mimiciv_icu.outputevents deINNER join  mimiciv_icu.d_items diON de.itemid = di.itemidINNER join  mimiciv_icu.icustays ieON de.stay_id = ie.stay_idWHERE de.subject_id = 10000980ORDER BY charttime;

"""
oe = pd.read_sql_query(query, conn)
# OPTION 2: load chartevents from a CSV file# ce = pd.read_csv('data/example_chartevents.csv', index_col='HOURSSINCEADMISSION')oe.head()print(oe.head())

数据量有点少，我们换一个患者10001725

为了给该图提供必要的背景，图中最好包含患者的输入数据，即inputevents表中的数据

这能为确定患者的体液平衡（患者健康的关键指标）提供了必要的参考依据。

请注意，列标题不同：我们有“HOURS_START”和“HOURS_END”。这是因为输入是在固定的时间内管理的。

列举label字段有哪些值？(重复的不显示)

这里我们取一个数值比较多的患者以方便后面输入输出对比 。de.subject_id = 10001884

import numpy as npimport pandas as pdimport matplotlib.pyplot as pltimport psycopg2
schema_name = 'mimic'# 连接到MIMIC-IV数据库conn = psycopg2.connect(dbname='mimiciv', user='postgres', password='mimic',                        host='10.241.148.228', port=5432)query_schema = 'SET search_path to ' + schema_name + ';'
# # # 设置查询语句# # # 我们选择从mimiciv_hosp.admissions表中提取hadm_id等于10006的行。# # # 在写sql代码时，最好先执行“set search_path to mimiciv" 随后的所有操作均不需要指明表格的位置；否则，任何操作都应该在表格名前面加前缀mimiciv# query1 = query_schema + 'SELECT subject_id, hadm_id, admittime, dischtime, admission_type FROM  mimiciv_hosp.admissions'## # 运行查询并将结果分配给变量# admissions_pd = pd.read_sql_query(query1,conn)# admissions_pd.head()# print(admissions_pd.head())
query = """

SELECT de.stay_id  , mimiciv_derived.datetime_diff(de.starttime, ie.intime, 'HOUR' ) AS HOURS_START    , mimiciv_derived.datetime_diff(de.endtime, ie.intime, 'HOUR' ) AS HOURS_END  , di.label    , de.amount    , de.amountuom    , de.rate    , de.rateuomFROM mimiciv_icu.inputevents deINNER join  mimiciv_icu.d_items diON de.itemid = di.itemidINNER join  mimiciv_icu.icustays ieON de.stay_id = ie.stay_idWHERE de.subject_id = 10001884ORDER BY endtime;

"""
ie = pd.read_sql_query(query, conn)


# OPTION 2: load chartevents from a CSV file# ce = pd.read_csv('data/example_chartevents.csv', index_col='HOURSSINCEADMISSION')ie.head()# print(ie.head())
ie['label'].unique()# print(ie['label'].unique())
ieml = ie[ie['amountuom'].str.contains('ml')]# print(ieml['hours_end'])# print(ieml['amount'])
query2 = """

SELECT de.stay_id  , mimiciv_derived.datetime_diff(de.charttime, ie.intime, 'HOUR' ) AS HOURS  , di.label  , de.value  , de.valueuomFROM mimiciv_icu.outputevents deINNER join  mimiciv_icu.d_items diON de.itemid = di.itemidINNER join  mimiciv_icu.icustays ieON de.stay_id = ie.stay_idWHERE de.subject_id = 10001884ORDER BY charttime;

"""
oe = pd.read_sql_query(query2, conn)
plt.figure(figsize=(14, 10))
# Plot the cumulative input against the cumulative outputplt.plot(ieml['hours_end'],         ieml['amount'].cumsum()/1000,         'go', markersize=8, label='Intake volume, L')
plt.plot(oe['hours'],         oe['value'].cumsum()/1000,         'ro', markersize=8, label='Output volume, L')
plt.title('Fluid balance over time',fontsize=16)plt.xlabel('Hours',fontsize=16)plt.ylabel('Volume (litres)',fontsize=16)# plt.ylim(0,38)plt.legend()plt.show()

如图所示，患者的摄入量往往高于其输出量（正如人们所期望的！），但在某些时期，它们几乎是一对一的。 

关注【科研收录】回复"患者输出体液"获取以上代码