Delete Mimic4Dataset.py
Browse files- Mimic4Dataset.py +0 -1178
Mimic4Dataset.py
DELETED
@@ -1,1178 +0,0 @@
|
|
1 |
-
import os
|
2 |
-
import pandas as pd
|
3 |
-
import datasets
|
4 |
-
import sys
|
5 |
-
import pickle
|
6 |
-
import subprocess
|
7 |
-
import shutil
|
8 |
-
from urllib.request import urlretrieve
|
9 |
-
from sklearn.model_selection import train_test_split
|
10 |
-
from sklearn.preprocessing import LabelEncoder
|
11 |
-
import numpy as np
|
12 |
-
from tqdm import tqdm
|
13 |
-
import yaml
|
14 |
-
import time
|
15 |
-
import torch
|
16 |
-
|
17 |
-
|
18 |
-
_DESCRIPTION = """\
|
19 |
-
Dataset for mimic4 data, by default for the Mortality task.
|
20 |
-
Available tasks are: Mortality, Length of Stay, Readmission, Phenotype.
|
21 |
-
The data is extracted from the mimic4 database using this pipeline: 'https://github.com/healthylaife/MIMIC-IV-Data-Pipeline/tree/main'
|
22 |
-
mimic path should have this form : "path/to/mimic4data/from/username/mimiciv/2.2"
|
23 |
-
If you choose a Custom task provide a configuration file for the Time series.
|
24 |
-
"""
|
25 |
-
|
26 |
-
_HOMEPAGE = "https://huggingface.co/datasets/thbndi/Mimic4Dataset"
|
27 |
-
_CITATION = "https://proceedings.mlr.press/v193/gupta22a.html"
|
28 |
-
_URL = "https://github.com/healthylaife/MIMIC-IV-Data-Pipeline"
|
29 |
-
_DATA_GEN = 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/data_generation_icu_modify.py'
|
30 |
-
_DATA_GEN_HOSP= 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/data_generation_modify.py'
|
31 |
-
_DAY_INT= 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/day_intervals_cohort_v22.py'
|
32 |
-
_CONFIG_URLS = {'los' : 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/config/los.config',
|
33 |
-
'mortality' : 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/config/mortality.config',
|
34 |
-
'phenotype' : 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/config/phenotype.config',
|
35 |
-
'readmission' : 'https://huggingface.co/datasets/thbndi/Mimic4Dataset/resolve/main/config/readmission.config'
|
36 |
-
}
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
def check_config(task,config_file):
|
41 |
-
with open(config_file) as f:
|
42 |
-
config = yaml.safe_load(f)
|
43 |
-
|
44 |
-
if task=='Phenotype':
|
45 |
-
disease_label = config['disease_label']
|
46 |
-
else :
|
47 |
-
disease_label = ""
|
48 |
-
time = config['timePrediction']
|
49 |
-
label = task
|
50 |
-
timeW = config['timeWindow']
|
51 |
-
include=int(timeW.split()[1])
|
52 |
-
bucket = config['timebucket']
|
53 |
-
radimp = config['radimp']
|
54 |
-
predW = config['predW']
|
55 |
-
disease_filter = config['disease_filter']
|
56 |
-
icu_no_icu = config['icu_no_icu']
|
57 |
-
groupingDiag = config['groupingDiag']
|
58 |
-
|
59 |
-
assert( icu_no_icu in ['ICU','Non-ICU' ], "Chossen data should be one of the following: ICU, Non-ICU")
|
60 |
-
data_icu = icu_no_icu=='ICU'
|
61 |
-
|
62 |
-
if data_icu:
|
63 |
-
chart_flag = config['chart']
|
64 |
-
output_flag = config['output']
|
65 |
-
select_chart = config['select_chart']
|
66 |
-
lab_flag = False
|
67 |
-
select_lab = False
|
68 |
-
else:
|
69 |
-
lab_flag =config['lab']
|
70 |
-
select_lab = config['select_lab']
|
71 |
-
groupingMed = config['groupingMed']
|
72 |
-
groupingProc = config['groupingProc']
|
73 |
-
chart_flag = False
|
74 |
-
output_flag = False
|
75 |
-
select_chart = False
|
76 |
-
|
77 |
-
|
78 |
-
diag_flag= config['diagnosis']
|
79 |
-
proc_flag = config['proc']
|
80 |
-
meds_flag = config['meds']
|
81 |
-
select_diag= config['select_diag']
|
82 |
-
select_med= config['select_med']
|
83 |
-
select_proc= config['select_proc']
|
84 |
-
select_out = config['select_out']
|
85 |
-
|
86 |
-
outlier_removal=config['outlier_removal']
|
87 |
-
thresh=config['outlier']
|
88 |
-
left_thresh=config['left_outlier']
|
89 |
-
|
90 |
-
if data_icu:
|
91 |
-
assert (isinstance(select_diag,bool) and isinstance(select_med,bool) and isinstance(select_proc,bool) and isinstance(select_out,bool) and isinstance(select_chart,bool), " select_diag, select_chart, select_med, select_proc, select_out should be boolean")
|
92 |
-
assert (isinstance(chart_flag,bool) and isinstance(output_flag,bool) and isinstance(diag_flag,bool) and isinstance(proc_flag,bool) and isinstance(meds_flag,bool), "chart_flag, output_flag, diag_flag, proc_flag, meds_flag should be boolean")
|
93 |
-
|
94 |
-
else:
|
95 |
-
assert (isinstance(select_diag,bool) and isinstance(select_med,bool) and isinstance(select_proc,bool) and isinstance(select_out,bool) and isinstance(select_lab,bool), " select_diag, select_lab, select_med, select_proc, select_out should be boolean")
|
96 |
-
assert (isinstance(lab_flag,bool) and isinstance(diag_flag,bool) and isinstance(proc_flag,bool) and isinstance(meds_flag,bool), "lab_flag, diag_flag, proc_flag, meds_flag should be boolean")
|
97 |
-
|
98 |
-
if task=='Phenotype':
|
99 |
-
if disease_label=='Heart Failure':
|
100 |
-
label='Readmission'
|
101 |
-
time=30
|
102 |
-
disease_label='I50'
|
103 |
-
elif disease_label=='CAD':
|
104 |
-
label='Readmission'
|
105 |
-
time=30
|
106 |
-
disease_label='I25'
|
107 |
-
elif disease_label=='CKD':
|
108 |
-
label='Readmission'
|
109 |
-
time=30
|
110 |
-
disease_label='N18'
|
111 |
-
elif disease_label=='COPD':
|
112 |
-
label='Readmission'
|
113 |
-
time=30
|
114 |
-
disease_label='J44'
|
115 |
-
else :
|
116 |
-
raise ValueError('Disease label not correct provide one in the list: Heart Failure, CAD, CKD, COPD')
|
117 |
-
predW=0
|
118 |
-
assert (timeW[0]=='Last' and include<=72 and include>=24, "Time window should be between Last 24 and Last 72")
|
119 |
-
|
120 |
-
elif task=='Mortality':
|
121 |
-
time=0
|
122 |
-
label= 'Mortality'
|
123 |
-
assert (predW<=8 and predW>=2, "Prediction window should be between 2 and 8")
|
124 |
-
assert (timeW[0]=='Fisrt' and include<=72 and include>=24, "Time window should be between First 24 and First 72")
|
125 |
-
|
126 |
-
elif task=='Length_of_Stay':
|
127 |
-
label= 'Length of Stay'
|
128 |
-
assert (timeW[0]=='Fisrt' and include<=72 and include>=24, "Time window should be between Fisrt 24 and Fisrt 72")
|
129 |
-
assert (time<=10 and time>=1, "Length of stay should be between 1 and 10")
|
130 |
-
predW=0
|
131 |
-
|
132 |
-
elif task=='Readmission':
|
133 |
-
label= 'Readmission'
|
134 |
-
assert (timeW[0]=='Last' and include<=72 and include>=24, "Time window should be between Last 24 and Last 72")
|
135 |
-
assert (time<=150 and time>=10 and time%10==0, "Readmission window should be between 10 and 150 with a step of 10")
|
136 |
-
predW=0
|
137 |
-
|
138 |
-
else:
|
139 |
-
raise ValueError('Task not correct')
|
140 |
-
|
141 |
-
assert( disease_filter in ['Heart Failure','COPD','CKD','CAD',""], "Disease filter should be one of the following: Heart Failure, COPD, CKD, CAD or empty")
|
142 |
-
assert( groupingDiag in ['Convert ICD-9 to ICD-10 and group ICD-10 codes','Keep both ICD-9 and ICD-10 codes','Convert ICD-9 to ICD-10 codes'], "Grouping ICD should be one of the following: Convert ICD-9 to ICD-10 and group ICD-10 codes, Keep both ICD-9 and ICD-10 codes, Convert ICD-9 to ICD-10 codes")
|
143 |
-
assert (bucket<=6 and bucket>=1 and isinstance(bucket, int), "Time bucket should be between 1 and 6 and an integer")
|
144 |
-
assert (radimp in ['No Imputation', 'forward fill and mean','forward fill and median'], "imputation should be one of the following: No Imputation, forward fill and mean, forward fill and median")
|
145 |
-
if chart_flag:
|
146 |
-
assert (left_thresh>=0 and left_thresh<=10 and isinstance(left_thresh, int), "Left outlier threshold should be between 0 and 10 and an integer")
|
147 |
-
assert (thresh>=90 and thresh<=99 and isinstance(thresh, int), "Outlier threshold should be between 90 and 99 and an integer")
|
148 |
-
assert (outlier_removal in ['No outlier detection','Impute Outlier (default:98)','Remove outliers (default:98)'], "Outlier removal should be one of the following: No outlier detection, Impute Outlier (default:98), Remove outliers (default:98)")
|
149 |
-
if lab_flag:
|
150 |
-
assert (left_thresh>=0 and left_thresh<=10 and isinstance(left_thresh, int), "Left outlier threshold should be between 0 and 10 and an integer")
|
151 |
-
assert (thresh>=90 and thresh<=99 and isinstance(thresh, int), "Outlier threshold should be between 90 and 99 and an integer")
|
152 |
-
assert (outlier_removal in ['No outlier detection','Impute Outlier (default:98)','Remove outliers (default:98)'], "Outlier removal should be one of the following: No outlier detection, Impute Outlier (default:98), Remove outliers (default:98)")
|
153 |
-
assert (groupingProc in ['ICD-9 and ICD-10','ICD-10'], "Grouping procedure should be one of the following: ICD-9 and ICD-10, ICD-10")
|
154 |
-
assert (groupingMed in ['Yes','No'], "Do you want to group Medication codes to use Non propietary names? : Grouping medication should be one of the following: Yes, No")
|
155 |
-
|
156 |
-
return label, time, disease_label, predW
|
157 |
-
|
158 |
-
def create_vocab(file,task):
|
159 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
160 |
-
condVocab = pickle.load(fp)
|
161 |
-
condVocabDict={}
|
162 |
-
condVocabDict[0]=0
|
163 |
-
for val in range(len(condVocab)):
|
164 |
-
condVocabDict[condVocab[val]]= val+1
|
165 |
-
|
166 |
-
return condVocabDict
|
167 |
-
|
168 |
-
def gender_vocab():
|
169 |
-
genderVocabDict={}
|
170 |
-
genderVocabDict['<PAD>']=0
|
171 |
-
genderVocabDict['M']=1
|
172 |
-
genderVocabDict['F']=2
|
173 |
-
|
174 |
-
return genderVocabDict
|
175 |
-
|
176 |
-
def vocab(task,diag_flag,proc_flag,out_flag,chart_flag,med_flag,lab_flag):
|
177 |
-
condVocabDict={}
|
178 |
-
procVocabDict={}
|
179 |
-
medVocabDict={}
|
180 |
-
outVocabDict={}
|
181 |
-
chartVocabDict={}
|
182 |
-
labVocabDict={}
|
183 |
-
ethVocabDict={}
|
184 |
-
ageVocabDict={}
|
185 |
-
genderVocabDict={}
|
186 |
-
insVocabDict={}
|
187 |
-
|
188 |
-
ethVocabDict=create_vocab('ethVocab',task)
|
189 |
-
with open('./data/dict/'+task+'/ethVocabDict', 'wb') as fp:
|
190 |
-
pickle.dump(ethVocabDict, fp)
|
191 |
-
|
192 |
-
ageVocabDict=create_vocab('ageVocab',task)
|
193 |
-
with open('./data/dict/'+task+'/ageVocabDict', 'wb') as fp:
|
194 |
-
pickle.dump(ageVocabDict, fp)
|
195 |
-
|
196 |
-
genderVocabDict=gender_vocab()
|
197 |
-
with open('./data/dict/'+task+'/genderVocabDict', 'wb') as fp:
|
198 |
-
pickle.dump(genderVocabDict, fp)
|
199 |
-
|
200 |
-
insVocabDict=create_vocab('insVocab',task)
|
201 |
-
with open('./data/dict/'+task+'/insVocabDict', 'wb') as fp:
|
202 |
-
pickle.dump(insVocabDict, fp)
|
203 |
-
|
204 |
-
if diag_flag:
|
205 |
-
file='condVocab'
|
206 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
207 |
-
condVocabDict = pickle.load(fp)
|
208 |
-
if proc_flag:
|
209 |
-
file='procVocab'
|
210 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
211 |
-
procVocabDict = pickle.load(fp)
|
212 |
-
if med_flag:
|
213 |
-
file='medVocab'
|
214 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
215 |
-
medVocabDict = pickle.load(fp)
|
216 |
-
if out_flag:
|
217 |
-
file='outVocab'
|
218 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
219 |
-
outVocabDict = pickle.load(fp)
|
220 |
-
if chart_flag:
|
221 |
-
file='chartVocab'
|
222 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
223 |
-
chartVocabDict = pickle.load(fp)
|
224 |
-
if lab_flag:
|
225 |
-
file='labsVocab'
|
226 |
-
with open ('./data/dict/'+task+'/'+file, 'rb') as fp:
|
227 |
-
labVocabDict = pickle.load(fp)
|
228 |
-
|
229 |
-
return len(condVocabDict),len(procVocabDict),len(medVocabDict),len(outVocabDict),len(chartVocabDict),len(labVocabDict),ethVocabDict,genderVocabDict,ageVocabDict,insVocabDict
|
230 |
-
|
231 |
-
def concat_data(data,task,feat_cond,feat_proc,feat_out,feat_chart,feat_meds,feat_lab):
|
232 |
-
meds=data['Med']
|
233 |
-
proc = data['Proc']
|
234 |
-
out = data['Out']
|
235 |
-
chart = data['Chart']
|
236 |
-
cond= data['Cond']['fids']
|
237 |
-
|
238 |
-
cond_df=pd.DataFrame()
|
239 |
-
proc_df=pd.DataFrame()
|
240 |
-
out_df=pd.DataFrame()
|
241 |
-
chart_df=pd.DataFrame()
|
242 |
-
meds_df=pd.DataFrame()
|
243 |
-
|
244 |
-
#demographic
|
245 |
-
demo=pd.DataFrame(columns=['Age','gender','ethnicity','label','insurance'])
|
246 |
-
new_row = {'Age': data['age'], 'gender': data['gender'], 'ethnicity': data['ethnicity'], 'label': data['label'], 'insurance': data['insurance']}
|
247 |
-
demo = demo.append(new_row, ignore_index=True)
|
248 |
-
|
249 |
-
##########COND#########
|
250 |
-
if (feat_cond):
|
251 |
-
#get all conds
|
252 |
-
with open("./data/dict/"+task+"/condVocab", 'rb') as fp:
|
253 |
-
conDict = pickle.load(fp)
|
254 |
-
conds=pd.DataFrame(conDict,columns=['COND'])
|
255 |
-
features=pd.DataFrame(np.zeros([1,len(conds)]),columns=conds['COND'])
|
256 |
-
|
257 |
-
#onehot encode
|
258 |
-
if(cond ==[]):
|
259 |
-
cond_df=pd.DataFrame(np.zeros([1,len(features)]),columns=features['COND'])
|
260 |
-
cond_df=cond_df.fillna(0)
|
261 |
-
else:
|
262 |
-
cond_df=pd.DataFrame(cond,columns=['COND'])
|
263 |
-
cond_df['val']=1
|
264 |
-
cond_df=(cond_df.drop_duplicates()).pivot(columns='COND',values='val').reset_index(drop=True)
|
265 |
-
cond_df=cond_df.fillna(0)
|
266 |
-
oneh = cond_df.sum().to_frame().T
|
267 |
-
combined_df = pd.concat([features,oneh],ignore_index=True).fillna(0)
|
268 |
-
combined_oneh=combined_df.sum().to_frame().T
|
269 |
-
cond_df=combined_oneh
|
270 |
-
|
271 |
-
##########PROC#########
|
272 |
-
if (feat_proc):
|
273 |
-
with open("./data/dict/"+task+"/procVocab", 'rb') as fp:
|
274 |
-
procDic = pickle.load(fp)
|
275 |
-
|
276 |
-
if proc :
|
277 |
-
feat=proc.keys()
|
278 |
-
proc_val=[proc[key] for key in feat]
|
279 |
-
procedures=pd.DataFrame(procDic,columns=['PROC'])
|
280 |
-
features=pd.DataFrame(np.zeros([1,len(procedures)]),columns=procedures['PROC'])
|
281 |
-
features.columns=pd.MultiIndex.from_product([["PROC"], features.columns])
|
282 |
-
procs=pd.DataFrame(columns=feat)
|
283 |
-
for p,v in zip(feat,proc_val):
|
284 |
-
procs[p]=v
|
285 |
-
procs.columns=pd.MultiIndex.from_product([["PROC"], procs.columns])
|
286 |
-
proc_df = pd.concat([features,procs],ignore_index=True).fillna(0)
|
287 |
-
else:
|
288 |
-
procedures=pd.DataFrame(procDic,columns=['PROC'])
|
289 |
-
features=pd.DataFrame(np.zeros([1,len(procedures)]),columns=procedures['PROC'])
|
290 |
-
features.columns=pd.MultiIndex.from_product([["PROC"], features.columns])
|
291 |
-
proc_df=features.fillna(0)
|
292 |
-
|
293 |
-
##########OUT#########
|
294 |
-
if (feat_out):
|
295 |
-
with open("./data/dict/"+task+"/outVocab", 'rb') as fp:
|
296 |
-
outDic = pickle.load(fp)
|
297 |
-
|
298 |
-
if out :
|
299 |
-
feat=out.keys()
|
300 |
-
out_val=[out[key] for key in feat]
|
301 |
-
outputs=pd.DataFrame(outDic,columns=['OUT'])
|
302 |
-
features=pd.DataFrame(np.zeros([1,len(outputs)]),columns=outputs['OUT'])
|
303 |
-
features.columns=pd.MultiIndex.from_product([["OUT"], features.columns])
|
304 |
-
outs=pd.DataFrame(columns=feat)
|
305 |
-
for o,v in zip(feat,out_val):
|
306 |
-
outs[o]=v
|
307 |
-
outs.columns=pd.MultiIndex.from_product([["OUT"], outs.columns])
|
308 |
-
out_df = pd.concat([features,outs],ignore_index=True).fillna(0)
|
309 |
-
else:
|
310 |
-
outputs=pd.DataFrame(outDic,columns=['OUT'])
|
311 |
-
features=pd.DataFrame(np.zeros([1,len(outputs)]),columns=outputs['OUT'])
|
312 |
-
features.columns=pd.MultiIndex.from_product([["OUT"], features.columns])
|
313 |
-
out_df=features.fillna(0)
|
314 |
-
|
315 |
-
##########CHART#########
|
316 |
-
if (feat_chart):
|
317 |
-
with open("./data/dict/"+task+"/chartVocab", 'rb') as fp:
|
318 |
-
chartDic = pickle.load(fp)
|
319 |
-
|
320 |
-
if chart:
|
321 |
-
charts=chart['val']
|
322 |
-
feat=charts.keys()
|
323 |
-
chart_val=[charts[key] for key in feat]
|
324 |
-
charts=pd.DataFrame(chartDic,columns=['CHART'])
|
325 |
-
features=pd.DataFrame(np.zeros([1,len(charts)]),columns=charts['CHART'])
|
326 |
-
features.columns=pd.MultiIndex.from_product([["CHART"], features.columns])
|
327 |
-
|
328 |
-
chart=pd.DataFrame(columns=feat)
|
329 |
-
for c,v in zip(feat,chart_val):
|
330 |
-
chart[c]=v
|
331 |
-
chart.columns=pd.MultiIndex.from_product([["CHART"], chart.columns])
|
332 |
-
chart_df = pd.concat([features,chart],ignore_index=True).fillna(0)
|
333 |
-
else:
|
334 |
-
charts=pd.DataFrame(chartDic,columns=['CHART'])
|
335 |
-
features=pd.DataFrame(np.zeros([1,len(charts)]),columns=charts['CHART'])
|
336 |
-
features.columns=pd.MultiIndex.from_product([["CHART"], features.columns])
|
337 |
-
chart_df=features.fillna(0)
|
338 |
-
|
339 |
-
##########LAB#########
|
340 |
-
if (feat_lab):
|
341 |
-
with open("./data/dict/"+task+"/labsVocab", 'rb') as fp:
|
342 |
-
chartDic = pickle.load(fp)
|
343 |
-
|
344 |
-
if chart:
|
345 |
-
charts=chart['val']
|
346 |
-
feat=charts.keys()
|
347 |
-
chart_val=[charts[key] for key in feat]
|
348 |
-
charts=pd.DataFrame(chartDic,columns=['LAB'])
|
349 |
-
features=pd.DataFrame(np.zeros([1,len(charts)]),columns=charts['LAB'])
|
350 |
-
features.columns=pd.MultiIndex.from_product([["LAB"], features.columns])
|
351 |
-
|
352 |
-
chart=pd.DataFrame(columns=feat)
|
353 |
-
for c,v in zip(feat,chart_val):
|
354 |
-
chart[c]=v
|
355 |
-
chart.columns=pd.MultiIndex.from_product([["LAB"], chart.columns])
|
356 |
-
chart_df = pd.concat([features,chart],ignore_index=True).fillna(0)
|
357 |
-
else:
|
358 |
-
charts=pd.DataFrame(chartDic,columns=['LAB'])
|
359 |
-
features=pd.DataFrame(np.zeros([1,len(charts)]),columns=charts['LAB'])
|
360 |
-
features.columns=pd.MultiIndex.from_product([["LAB"], features.columns])
|
361 |
-
chart_df=features.fillna(0)
|
362 |
-
|
363 |
-
###MEDS
|
364 |
-
if (feat_meds):
|
365 |
-
with open("./data/dict/"+task+"/medVocab", 'rb') as fp:
|
366 |
-
medDic = pickle.load(fp)
|
367 |
-
|
368 |
-
if meds:
|
369 |
-
feat=meds['signal'].keys()
|
370 |
-
med_val=[meds['amount'][key] for key in feat]
|
371 |
-
meds=pd.DataFrame(medDic,columns=['MEDS'])
|
372 |
-
features=pd.DataFrame(np.zeros([1,len(meds)]),columns=meds['MEDS'])
|
373 |
-
features.columns=pd.MultiIndex.from_product([["MEDS"], features.columns])
|
374 |
-
|
375 |
-
med=pd.DataFrame(columns=feat)
|
376 |
-
for m,v in zip(feat,med_val):
|
377 |
-
med[m]=v
|
378 |
-
med.columns=pd.MultiIndex.from_product([["MEDS"], med.columns])
|
379 |
-
meds_df = pd.concat([features,med],ignore_index=True).fillna(0)
|
380 |
-
else:
|
381 |
-
meds=pd.DataFrame(medDic,columns=['MEDS'])
|
382 |
-
features=pd.DataFrame(np.zeros([1,len(meds)]),columns=meds['MEDS'])
|
383 |
-
features.columns=pd.MultiIndex.from_product([["MEDS"], features.columns])
|
384 |
-
meds_df=features.fillna(0)
|
385 |
-
|
386 |
-
dyn_df = pd.concat([meds_df,proc_df,out_df,chart_df], axis=1)
|
387 |
-
return dyn_df,cond_df,demo
|
388 |
-
def getXY_deep(data,task,feat_cond,feat_proc,feat_out,feat_chart,feat_meds,feat_lab):
|
389 |
-
stat_df = torch.zeros(size=(1,0))
|
390 |
-
demo_df = torch.zeros(size=(1,0))
|
391 |
-
meds = torch.zeros(size=(0,0))
|
392 |
-
charts = torch.zeros(size=(0,0))
|
393 |
-
proc = torch.zeros(size=(0,0))
|
394 |
-
out = torch.zeros(size=(0,0))
|
395 |
-
lab = torch.zeros(size=(0,0))
|
396 |
-
stat_df = torch.zeros(size=(1,0))
|
397 |
-
demo_df = torch.zeros(size=(1,0))
|
398 |
-
|
399 |
-
size_cond, size_proc, size_meds, size_out, size_chart, size_lab, eth_vocab,gender_vocab,age_vocab,ins_vocab=vocab(task.replace(" ","_"),feat_cond,feat_proc,feat_out,feat_chart,feat_meds,False)
|
400 |
-
dyn,cond_df,demo=concat_data(data,task.replace(" ","_"),feat_cond,feat_proc,feat_out,feat_chart,feat_meds,feat_lab)
|
401 |
-
|
402 |
-
###########""
|
403 |
-
if feat_chart:
|
404 |
-
charts = dyn['CHART']
|
405 |
-
charts=charts.to_numpy()
|
406 |
-
charts = torch.tensor(charts, dtype=torch.long)
|
407 |
-
charts = charts.tolist()
|
408 |
-
|
409 |
-
if feat_meds:
|
410 |
-
meds = dyn['MEDS']
|
411 |
-
meds=meds.to_numpy()
|
412 |
-
meds = torch.tensor(meds, dtype=torch.long)
|
413 |
-
meds = meds.tolist()
|
414 |
-
|
415 |
-
if feat_proc:
|
416 |
-
proc = dyn['PROC']
|
417 |
-
proc=proc.to_numpy()
|
418 |
-
proc = torch.tensor(proc, dtype=torch.long)
|
419 |
-
proc = proc.tolist()
|
420 |
-
|
421 |
-
if feat_out:
|
422 |
-
out = dyn['OUT']
|
423 |
-
out=out.to_numpy()
|
424 |
-
out = torch.tensor(out, dtype=torch.long)
|
425 |
-
out = out.tolist()
|
426 |
-
|
427 |
-
if feat_lab:
|
428 |
-
lab = dyn['LAB']
|
429 |
-
lab=lab.to_numpy()
|
430 |
-
lab = torch.tensor(lab, dtype=torch.long)
|
431 |
-
lab = lab.tolist()
|
432 |
-
|
433 |
-
####################""
|
434 |
-
|
435 |
-
stat=cond_df
|
436 |
-
stat = stat.to_numpy()
|
437 |
-
stat = torch.tensor(stat)
|
438 |
-
if stat_df[0].nelement():
|
439 |
-
stat_df = torch.cat((stat_df,stat),0)
|
440 |
-
else:
|
441 |
-
stat_df = stat
|
442 |
-
|
443 |
-
y = int(demo['label'])
|
444 |
-
demo["gender"].replace(gender_vocab, inplace=True)
|
445 |
-
demo["ethnicity"].replace(eth_vocab, inplace=True)
|
446 |
-
demo["insurance"].replace(ins_vocab, inplace=True)
|
447 |
-
demo["Age"].replace(age_vocab, inplace=True)
|
448 |
-
demo=demo[["gender","ethnicity","insurance","Age"]]
|
449 |
-
demo = demo.values
|
450 |
-
demo = torch.tensor(demo)
|
451 |
-
if demo_df[0].nelement():
|
452 |
-
demo_df = torch.cat((demo_df,demo),0)
|
453 |
-
else:
|
454 |
-
demo_df = demo
|
455 |
-
stat_df = torch.tensor(stat_df)
|
456 |
-
stat_df = stat_df.type(torch.LongTensor)
|
457 |
-
stat_df = stat_df.squeeze()
|
458 |
-
demo_df = torch.tensor(demo_df)
|
459 |
-
demo_df = demo_df.type(torch.LongTensor)
|
460 |
-
demo_df=demo_df.squeeze()
|
461 |
-
y_df = torch.tensor(y)
|
462 |
-
y_df = y_df.type(torch.LongTensor)
|
463 |
-
|
464 |
-
return stat_df, demo_df, meds, charts, out, proc, lab, y
|
465 |
-
|
466 |
-
def getXY(dyn,stat,demo,concat_cols,concat):
|
467 |
-
X_df=pd.DataFrame()
|
468 |
-
if concat:
|
469 |
-
dyna=dyn.copy()
|
470 |
-
dyna.columns=dyna.columns.droplevel(0)
|
471 |
-
dyna=dyna.to_numpy()
|
472 |
-
dyna=np.nan_to_num(dyna, copy=False)
|
473 |
-
dyna=dyna.reshape(1,-1)
|
474 |
-
dyn_df=pd.DataFrame(data=dyna,columns=concat_cols)
|
475 |
-
else:
|
476 |
-
dyn_df=pd.DataFrame()
|
477 |
-
for key in dyn.columns.levels[0]:
|
478 |
-
dyn_temp=dyn[key]
|
479 |
-
if ((key=="CHART") or (key=="MEDS")):
|
480 |
-
agg=dyn_temp.aggregate("mean")
|
481 |
-
agg=agg.reset_index()
|
482 |
-
else:
|
483 |
-
agg=dyn_temp.aggregate("max")
|
484 |
-
agg=agg.reset_index()
|
485 |
-
|
486 |
-
if dyn_df.empty:
|
487 |
-
dyn_df=agg
|
488 |
-
else:
|
489 |
-
dyn_df=pd.concat([dyn_df,agg],axis=0)
|
490 |
-
dyn_df=dyn_df.T
|
491 |
-
dyn_df.columns = dyn_df.iloc[0]
|
492 |
-
dyn_df=dyn_df.iloc[1:,:]
|
493 |
-
|
494 |
-
X_df=pd.concat([dyn_df,stat],axis=1)
|
495 |
-
X_df=pd.concat([X_df,demo],axis=1)
|
496 |
-
return X_df
|
497 |
-
|
498 |
-
|
499 |
-
|
500 |
-
|
501 |
-
|
502 |
-
|
503 |
-
def task_cohort(task, mimic_path, config_path):
|
504 |
-
sys.path.append('./preprocessing/day_intervals_preproc')
|
505 |
-
sys.path.append('./utils')
|
506 |
-
sys.path.append('./preprocessing/hosp_module_preproc')
|
507 |
-
sys.path.append('./model')
|
508 |
-
import day_intervals_cohort_v22
|
509 |
-
import day_intervals_cohort
|
510 |
-
import feature_selection_icu
|
511 |
-
import data_generation_icu_modify
|
512 |
-
import data_generation_modify
|
513 |
-
import feature_selection_hosp
|
514 |
-
|
515 |
-
|
516 |
-
root_dir = os.path.dirname(os.path.abspath('UserInterface.ipynb'))
|
517 |
-
config_path='./config/'+config_path
|
518 |
-
with open(config_path) as f:
|
519 |
-
config = yaml.safe_load(f)
|
520 |
-
version_path = mimic_path+'/'
|
521 |
-
print(version_path)
|
522 |
-
version = mimic_path.split('/')[-1][0]
|
523 |
-
start = time.time()
|
524 |
-
#----------------------------------------------config----------------------------------------------------
|
525 |
-
label, tim, disease_label, predW = check_config(task,config_path)
|
526 |
-
icu_no_icu = config['icu_no_icu']
|
527 |
-
timeW = config['timeWindow']
|
528 |
-
include=int(timeW.split()[1])
|
529 |
-
bucket = config['timebucket']
|
530 |
-
radimp = config['radimp']
|
531 |
-
|
532 |
-
diag_flag = config['diagnosis']
|
533 |
-
proc_flag= config['proc']
|
534 |
-
med_flag = config['meds']
|
535 |
-
disease_filter = config['disease_filter']
|
536 |
-
groupingDiag = config['groupingDiag']
|
537 |
-
select_diag= config['select_diag']
|
538 |
-
select_med= config['select_med']
|
539 |
-
select_proc= config['select_proc']
|
540 |
-
|
541 |
-
if icu_no_icu=='ICU':
|
542 |
-
out_flag = config['output']
|
543 |
-
chart_flag = config['chart']
|
544 |
-
select_out= config['select_out']
|
545 |
-
select_chart= config['select_chart']
|
546 |
-
lab_flag = False
|
547 |
-
select_lab = False
|
548 |
-
else:
|
549 |
-
lab_flag = config['lab']
|
550 |
-
groupingMed = config['groupingMed']
|
551 |
-
groupingProc = config['groupingProc']
|
552 |
-
select_lab= config['select_lab']
|
553 |
-
out_flag = False
|
554 |
-
chart_flag = False
|
555 |
-
select_out= False
|
556 |
-
select_chart= False
|
557 |
-
|
558 |
-
# -------------------------------------------------------------------------------------------------------------
|
559 |
-
|
560 |
-
data_icu=icu_no_icu=="ICU"
|
561 |
-
data_mort=label=="Mortality"
|
562 |
-
data_admn=label=='Readmission'
|
563 |
-
data_los=label=='Length of Stay'
|
564 |
-
|
565 |
-
if (disease_filter=="Heart Failure"):
|
566 |
-
icd_code='I50'
|
567 |
-
elif (disease_filter=="CKD"):
|
568 |
-
icd_code='N18'
|
569 |
-
elif (disease_filter=="COPD"):
|
570 |
-
icd_code='J44'
|
571 |
-
elif (disease_filter=="CAD"):
|
572 |
-
icd_code='I25'
|
573 |
-
else:
|
574 |
-
icd_code='No Disease Filter'
|
575 |
-
|
576 |
-
#-----------------------------------------------EXTRACT MIMIC-----------------------------------------------------
|
577 |
-
if version == '2':
|
578 |
-
cohort_output = day_intervals_cohort_v22.extract_data(icu_no_icu,label,tim,icd_code, root_dir,version_path,disease_label)
|
579 |
-
|
580 |
-
elif version == '1':
|
581 |
-
cohort_output = day_intervals_cohort.extract_data(icu_no_icu,label,tim,icd_code, root_dir,version_path,disease_label)
|
582 |
-
#----------------------------------------------FEATURES-------------------------------------------------------
|
583 |
-
|
584 |
-
if data_icu :
|
585 |
-
feature_selection_icu.feature_icu(cohort_output, version_path,diag_flag,out_flag,chart_flag,proc_flag,med_flag)
|
586 |
-
else:
|
587 |
-
feature_selection_hosp.feature_nonicu(cohort_output, version_path,diag_flag,lab_flag,proc_flag,med_flag)
|
588 |
-
#----------------------------------------------GROUPING-------------------------------------------------------
|
589 |
-
if data_icu:
|
590 |
-
if diag_flag:
|
591 |
-
group_diag=groupingDiag
|
592 |
-
feature_selection_icu.preprocess_features_icu(cohort_output, diag_flag, group_diag,False,False,False,0,0)
|
593 |
-
|
594 |
-
else:
|
595 |
-
if diag_flag:
|
596 |
-
group_diag=groupingDiag
|
597 |
-
if med_flag:
|
598 |
-
group_med=groupingMed
|
599 |
-
if proc_flag:
|
600 |
-
group_proc=groupingProc
|
601 |
-
feature_selection_hosp.preprocess_features_hosp(cohort_output, diag_flag,proc_flag,med_flag,False,group_diag,group_med,group_proc,False,False,0,0)
|
602 |
-
#----------------------------------------------SUMMARY-------------------------------------------------------
|
603 |
-
if data_icu:
|
604 |
-
feature_selection_icu.generate_summary_icu(diag_flag,proc_flag,med_flag,out_flag,chart_flag)
|
605 |
-
else:
|
606 |
-
feature_selection_hosp.generate_summary_hosp(diag_flag,proc_flag,med_flag,lab_flag)
|
607 |
-
#----------------------------------------------FEATURE SELECTION---------------------------------------------
|
608 |
-
|
609 |
-
#----------------------------------------------FEATURE SELECTION---------------------------------------------
|
610 |
-
|
611 |
-
if data_icu:
|
612 |
-
if select_chart or select_out or select_diag or select_med or select_proc:
|
613 |
-
if select_chart:
|
614 |
-
input('Please edit list of codes in ./data/summary/chart_features.csv to select the chart items to keep and press enter to continue')
|
615 |
-
if select_out:
|
616 |
-
input('Please edit list of codes in ./data/summary/out_features.csv to select the output items to keep and press enter to continue')
|
617 |
-
if select_diag:
|
618 |
-
input('Please edit list of codes in ./data/summary/diag_features.csv to select the diagnosis ids to keep and press enter to continue')
|
619 |
-
if select_med:
|
620 |
-
input('Please edit list of codes in ./data/summary/med_features.csv to select the meds items to keep and press enter to continue')
|
621 |
-
if select_proc:
|
622 |
-
input('Please edit list of codes in ./data/summary/proc_features.csv to select the procedures ids to keep and press enter to continue')
|
623 |
-
feature_selection_icu.features_selection_icu(cohort_output, diag_flag,proc_flag,med_flag,out_flag, chart_flag,select_diag,select_med,select_proc,select_out,select_chart)
|
624 |
-
else:
|
625 |
-
if select_diag or select_med or select_proc or select_lab:
|
626 |
-
if select_diag:
|
627 |
-
input('Please edit list of codes in ./data/summary/diag_features.csv to select the diagnosis ids to keep and press enter to continue')
|
628 |
-
if select_med:
|
629 |
-
input('Please edit list of codes in ./data/summary/med_features.csv to select the meds items to keep and press enter to continue')
|
630 |
-
if select_proc:
|
631 |
-
input('Please edit list of codes in ./data/summary/proc_features.csv to select the procedures ids to keep and press enter to continue')
|
632 |
-
if select_lab:
|
633 |
-
input('Please edit list of codes in ./data/summary/labs_features.csv to select the labs items to keep and press enter to continue')
|
634 |
-
feature_selection_hosp.features_selection_hosp(cohort_output, diag_flag,proc_flag,med_flag,lab_flag,select_diag,select_med,select_proc,select_lab)
|
635 |
-
|
636 |
-
#---------------------------------------CLEANING OF FEATURES-----------------------------------------------
|
637 |
-
thresh=0
|
638 |
-
if data_icu:
|
639 |
-
if chart_flag:
|
640 |
-
outlier_removal=config['outlier_removal']
|
641 |
-
clean_chart=outlier_removal!='No outlier detection'
|
642 |
-
impute_outlier_chart=outlier_removal=='Impute Outlier (default:98)'
|
643 |
-
thresh=config['outlier']
|
644 |
-
left_thresh=config['left_outlier']
|
645 |
-
feature_selection_icu.preprocess_features_icu(cohort_output, False, False,chart_flag,clean_chart,impute_outlier_chart,thresh,left_thresh)
|
646 |
-
else:
|
647 |
-
if lab_flag:
|
648 |
-
outlier_removal=config['outlier_removal']
|
649 |
-
clean_chart=outlier_removal!='No outlier detection'
|
650 |
-
impute_outlier_chart=outlier_removal=='Impute Outlier (default:98)'
|
651 |
-
thresh=config['outlier']
|
652 |
-
left_thresh=config['left_outlier']
|
653 |
-
feature_selection_hosp.preprocess_features_hosp(cohort_output, False,False, False,lab_flag,False,False,False,clean_chart,impute_outlier_chart,thresh,left_thresh)
|
654 |
-
# ---------------------------------------tim-Series Representation--------------------------------------------
|
655 |
-
if radimp == 'forward fill and mean' :
|
656 |
-
impute='Mean'
|
657 |
-
elif radimp =='forward fill and median':
|
658 |
-
impute = 'Median'
|
659 |
-
else :
|
660 |
-
impute = False
|
661 |
-
|
662 |
-
if data_icu:
|
663 |
-
gen=data_generation_icu_modify.Generator(task,cohort_output,data_mort,data_admn,data_los,diag_flag,proc_flag,out_flag,chart_flag,med_flag,impute,include,bucket,predW)
|
664 |
-
else:
|
665 |
-
gen=data_generation_modify.Generator(cohort_output,data_mort,data_admn,data_los,diag_flag,lab_flag,proc_flag,med_flag,impute,include,bucket,predW)
|
666 |
-
|
667 |
-
end = time.time()
|
668 |
-
print("Time elapsed : ", round((end - start)/60,2),"mins")
|
669 |
-
print("[============TASK COHORT SUCCESSFULLY CREATED============]")
|
670 |
-
|
671 |
-
|
672 |
-
|
673 |
-
|
674 |
-
#############################################DATASET####################################################################
|
675 |
-
class Mimic4DatasetConfig(datasets.BuilderConfig):
|
676 |
-
"""BuilderConfig for Mimic4Dataset."""
|
677 |
-
|
678 |
-
def __init__(
|
679 |
-
self,
|
680 |
-
**kwargs,
|
681 |
-
):
|
682 |
-
super().__init__(**kwargs)
|
683 |
-
|
684 |
-
class Mimic4Dataset(datasets.GeneratorBasedBuilder):
|
685 |
-
VERSION = datasets.Version("1.0.0")
|
686 |
-
|
687 |
-
def __init__(self, **kwargs):
|
688 |
-
self.mimic_path = kwargs.pop("mimic_path", None)
|
689 |
-
self.encoding = kwargs.pop("encoding",'raw')
|
690 |
-
self.config_path = kwargs.pop("config_path",None)
|
691 |
-
self.test_size = kwargs.pop("test_size",0.2)
|
692 |
-
self.val_size = kwargs.pop("val_size",0.1)
|
693 |
-
self.generate_cohort = kwargs.pop("generate_cohort",True)
|
694 |
-
|
695 |
-
if self.encoding == 'concat':
|
696 |
-
self.concat = True
|
697 |
-
else:
|
698 |
-
self.concat = False
|
699 |
-
|
700 |
-
super().__init__(**kwargs)
|
701 |
-
|
702 |
-
|
703 |
-
BUILDER_CONFIGS = [
|
704 |
-
Mimic4DatasetConfig(
|
705 |
-
name="Phenotype",
|
706 |
-
version=VERSION,
|
707 |
-
description="Dataset for mimic4 Phenotype task"
|
708 |
-
),
|
709 |
-
Mimic4DatasetConfig(
|
710 |
-
name="Readmission",
|
711 |
-
version=VERSION,
|
712 |
-
description="Dataset for mimic4 Readmission task"
|
713 |
-
),
|
714 |
-
Mimic4DatasetConfig(
|
715 |
-
name="Length of Stay",
|
716 |
-
version=VERSION,
|
717 |
-
description="Dataset for mimic4 Length of Stay task"
|
718 |
-
),
|
719 |
-
Mimic4DatasetConfig(
|
720 |
-
name="Mortality",
|
721 |
-
version=VERSION,
|
722 |
-
description="Dataset for mimic4 Mortality task"
|
723 |
-
),
|
724 |
-
]
|
725 |
-
|
726 |
-
DEFAULT_CONFIG_NAME = "Mortality"
|
727 |
-
|
728 |
-
def create_cohort(self):
|
729 |
-
if self.config_path==None:
|
730 |
-
if self.config.name == 'Phenotype' : self.config_path = _CONFIG_URLS['phenotype']
|
731 |
-
if self.config.name == 'Readmission' : self.config_path = _CONFIG_URLS['readmission']
|
732 |
-
if self.config.name == 'Length of Stay' : self.config_path = _CONFIG_URLS['los']
|
733 |
-
if self.config.name == 'Mortality' : self.config_path = _CONFIG_URLS['mortality']
|
734 |
-
|
735 |
-
version = self.mimic_path.split('/')[-1]
|
736 |
-
mimic_folder= self.mimic_path.split('/')[-2]
|
737 |
-
mimic_complete_path='/'+mimic_folder+'/'+version
|
738 |
-
|
739 |
-
current_directory = os.getcwd()
|
740 |
-
if os.path.exists(os.path.dirname(current_directory)+'/MIMIC-IV-Data-Pipeline-main'):
|
741 |
-
dir =os.path.dirname(current_directory)
|
742 |
-
os.chdir(dir)
|
743 |
-
else:
|
744 |
-
#move to parent directory of mimic data
|
745 |
-
dir = self.mimic_path.replace(mimic_complete_path,'')
|
746 |
-
if dir[-1]!='/':
|
747 |
-
dir=dir+'/'
|
748 |
-
elif dir=='':
|
749 |
-
dir="./"
|
750 |
-
parent_dir = os.path.dirname(self.mimic_path)
|
751 |
-
os.chdir(parent_dir)
|
752 |
-
|
753 |
-
#####################clone git repo if doesnt exists
|
754 |
-
repo_url='https://github.com/healthylaife/MIMIC-IV-Data-Pipeline'
|
755 |
-
if os.path.exists('MIMIC-IV-Data-Pipeline-main'):
|
756 |
-
path_bench = './MIMIC-IV-Data-Pipeline-main'
|
757 |
-
else:
|
758 |
-
path_bench ='./MIMIC-IV-Data-Pipeline-main'
|
759 |
-
subprocess.run(["git", "clone", repo_url, path_bench])
|
760 |
-
os.makedirs(path_bench+'/mimic-iv')
|
761 |
-
shutil.move(version,path_bench+'/mimic-iv')
|
762 |
-
|
763 |
-
os.chdir(path_bench)
|
764 |
-
self.mimic_path = './mimic-iv/'+version
|
765 |
-
|
766 |
-
####################Get configurations param
|
767 |
-
#download config file if not custom
|
768 |
-
if self.config_path[0:4] == 'http':
|
769 |
-
c = self.config_path.split('/')[-1]
|
770 |
-
file_path, head = urlretrieve(self.config_path,c)
|
771 |
-
else :
|
772 |
-
file_path = self.config_path
|
773 |
-
|
774 |
-
if not os.path.exists('./config'):
|
775 |
-
os.makedirs('config')
|
776 |
-
#save config file in config folder
|
777 |
-
self.conf='./config/'+file_path.split('/')[-1]
|
778 |
-
if not os.path.exists(self.conf):
|
779 |
-
shutil.move(file_path,'./config')
|
780 |
-
with open(self.conf) as f:
|
781 |
-
config = yaml.safe_load(f)
|
782 |
-
timeW = config['timeWindow']
|
783 |
-
self.timeW=int(timeW.split()[1])
|
784 |
-
self.bucket = config['timebucket']
|
785 |
-
self.data_icu = config['icu_no_icu']=='ICU'
|
786 |
-
if self.data_icu:
|
787 |
-
self.feat_cond, self.feat_chart, self.feat_proc, self.feat_meds, self.feat_out, self.lab = config['diagnosis'], config['chart'], config['proc'], config['meds'], config['output'], False
|
788 |
-
self.feat_lab = False
|
789 |
-
else:
|
790 |
-
self.feat_cond, self.feat_lab, self.feat_proc, self.feat_meds, self.feat_chart, self.out = config['diagnosis'], config['lab'], config['proc'], config['meds'], False, False
|
791 |
-
self.feat_out = False
|
792 |
-
self.feat_chart = False
|
793 |
-
#####################downloads modules from hub
|
794 |
-
if not os.path.exists('./model/data_generation_icu_modify.py'):
|
795 |
-
file_path, head = urlretrieve(_DATA_GEN, "data_generation_icu_modify.py")
|
796 |
-
shutil.move(file_path, './model')
|
797 |
-
|
798 |
-
if not os.path.exists('./model/data_generation_modify.py'):
|
799 |
-
file_path, head = urlretrieve(_DATA_GEN_HOSP, "data_generation_modify.py")
|
800 |
-
shutil.move(file_path, './model')
|
801 |
-
|
802 |
-
if not os.path.exists('./preprocessing/day_intervals_preproc/day_intervals_cohort_v22.py'):
|
803 |
-
file_path, head = urlretrieve(_DAY_INT, "day_intervals_cohort_v22.py")
|
804 |
-
shutil.move(file_path, './preprocessing/day_intervals_preproc')
|
805 |
-
|
806 |
-
data_dir = "./data/dict/"+self.config.name.replace(" ","_")+"/dataDic"
|
807 |
-
sys.path.append(path_bench)
|
808 |
-
config = self.config_path.split('/')[-1]
|
809 |
-
|
810 |
-
#####################create task cohort
|
811 |
-
if self.generate_cohort:
|
812 |
-
task_cohort(self.config.name.replace(" ","_"),self.mimic_path,config)
|
813 |
-
|
814 |
-
#####################Split data into train, test and val
|
815 |
-
with open(data_dir, 'rb') as fp:
|
816 |
-
dataDic = pickle.load(fp)
|
817 |
-
data = pd.DataFrame.from_dict(dataDic)
|
818 |
-
|
819 |
-
dict_dir = "./data/dict/"+self.config.name.replace(" ","_")
|
820 |
-
|
821 |
-
data=data.T
|
822 |
-
train_data, test_data = train_test_split(data, test_size=self.test_size, random_state=42)
|
823 |
-
if self.val_size > 0 :
|
824 |
-
train_data, val_data = train_test_split(train_data, test_size=self.val_size, random_state=42)
|
825 |
-
val_dic = val_data.to_dict('index')
|
826 |
-
val_path = dict_dir+'/val_data.pkl'
|
827 |
-
with open(val_path, 'wb') as f:
|
828 |
-
pickle.dump(val_dic, f)
|
829 |
-
|
830 |
-
train_dic = train_data.to_dict('index')
|
831 |
-
test_dic = test_data.to_dict('index')
|
832 |
-
|
833 |
-
train_path = dict_dir+'/train_data.pkl'
|
834 |
-
test_path = dict_dir+'/test_data.pkl'
|
835 |
-
|
836 |
-
|
837 |
-
with open(train_path, 'wb') as f:
|
838 |
-
pickle.dump(train_dic, f)
|
839 |
-
|
840 |
-
with open(test_path, 'wb') as f:
|
841 |
-
pickle.dump(test_dic, f)
|
842 |
-
|
843 |
-
return dict_dir
|
844 |
-
|
845 |
-
###########################################################RAW##################################################################
|
846 |
-
|
847 |
-
def _info_raw(self):
|
848 |
-
features = datasets.Features(
|
849 |
-
{
|
850 |
-
"label": datasets.ClassLabel(num_classes=2,names=["0", "1"]),
|
851 |
-
"gender": datasets.Value("string"),
|
852 |
-
"ethnicity": datasets.Value("string"),
|
853 |
-
"insurance": datasets.Value("string"),
|
854 |
-
"age": datasets.Value("int32"),
|
855 |
-
"COND": datasets.Sequence(datasets.Value("string")),
|
856 |
-
"MEDS": {
|
857 |
-
"signal":
|
858 |
-
{
|
859 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
860 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
861 |
-
}
|
862 |
-
,
|
863 |
-
"rate":
|
864 |
-
{
|
865 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
866 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
867 |
-
}
|
868 |
-
,
|
869 |
-
"amount":
|
870 |
-
{
|
871 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
872 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
873 |
-
}
|
874 |
-
|
875 |
-
},
|
876 |
-
"PROC": {
|
877 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
878 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
879 |
-
},
|
880 |
-
"CHART/LAB":
|
881 |
-
{
|
882 |
-
"signal" : {
|
883 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
884 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
885 |
-
},
|
886 |
-
"val" : {
|
887 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
888 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
889 |
-
},
|
890 |
-
},
|
891 |
-
"OUT": {
|
892 |
-
"id": datasets.Sequence(datasets.Value("int32")),
|
893 |
-
"value": datasets.Sequence(datasets.Sequence(datasets.Value("float32")))
|
894 |
-
},
|
895 |
-
|
896 |
-
}
|
897 |
-
)
|
898 |
-
return datasets.DatasetInfo(
|
899 |
-
description=_DESCRIPTION,
|
900 |
-
features=features,
|
901 |
-
homepage=_HOMEPAGE,
|
902 |
-
citation=_CITATION,
|
903 |
-
)
|
904 |
-
|
905 |
-
def _generate_examples_raw(self, filepath):
|
906 |
-
with open(filepath, 'rb') as fp:
|
907 |
-
dataDic = pickle.load(fp)
|
908 |
-
for hid, data in dataDic.items():
|
909 |
-
proc_features = data['Proc']
|
910 |
-
meds_features = data['Med']
|
911 |
-
out_features = data['Out']
|
912 |
-
cond_features = data['Cond']['fids']
|
913 |
-
eth= data['ethnicity']
|
914 |
-
age = data['age']
|
915 |
-
gender = data['gender']
|
916 |
-
label = data['label']
|
917 |
-
insurance=data['insurance']
|
918 |
-
|
919 |
-
items = list(proc_features.keys())
|
920 |
-
values =[proc_features[i] for i in items ]
|
921 |
-
procs = {"id" : items,
|
922 |
-
"value": values}
|
923 |
-
|
924 |
-
items_outs = list(out_features.keys())
|
925 |
-
values_outs =[out_features[i] for i in items_outs ]
|
926 |
-
outs = {"id" : items_outs,
|
927 |
-
"value": values_outs}
|
928 |
-
|
929 |
-
if self.data_icu:
|
930 |
-
chart_features = data['Chart']
|
931 |
-
else:
|
932 |
-
chart_features = data['Lab']
|
933 |
-
|
934 |
-
#chart signal
|
935 |
-
if ('signal' in chart_features):
|
936 |
-
items_chart_sig = list(chart_features['signal'].keys())
|
937 |
-
values_chart_sig =[chart_features['signal'][i] for i in items_chart_sig ]
|
938 |
-
chart_sig = {"id" : items_chart_sig,
|
939 |
-
"value": values_chart_sig}
|
940 |
-
else:
|
941 |
-
chart_sig = {"id" : [],
|
942 |
-
"value": []}
|
943 |
-
#chart val
|
944 |
-
if ('val' in chart_features):
|
945 |
-
items_chart_val = list(chart_features['val'].keys())
|
946 |
-
values_chart_val =[chart_features['val'][i] for i in items_chart_val ]
|
947 |
-
chart_val = {"id" : items_chart_val,
|
948 |
-
"value": values_chart_val}
|
949 |
-
else:
|
950 |
-
chart_val = {"id" : [],
|
951 |
-
"value": []}
|
952 |
-
|
953 |
-
charts = {"signal" : chart_sig,
|
954 |
-
"val" : chart_val}
|
955 |
-
|
956 |
-
#meds signal
|
957 |
-
if ('signal' in meds_features):
|
958 |
-
items_meds_sig = list(meds_features['signal'].keys())
|
959 |
-
values_meds_sig =[meds_features['signal'][i] for i in items_meds_sig ]
|
960 |
-
meds_sig = {"id" : items_meds_sig,
|
961 |
-
"value": values_meds_sig}
|
962 |
-
else:
|
963 |
-
meds_sig = {"id" : [],
|
964 |
-
"value": []}
|
965 |
-
#meds rate
|
966 |
-
if ('rate' in meds_features):
|
967 |
-
items_meds_rate = list(meds_features['rate'].keys())
|
968 |
-
values_meds_rate =[meds_features['rate'][i] for i in items_meds_rate ]
|
969 |
-
meds_rate = {"id" : items_meds_rate,
|
970 |
-
"value": values_meds_rate}
|
971 |
-
else:
|
972 |
-
meds_rate = {"id" : [],
|
973 |
-
"value": []}
|
974 |
-
#meds amount
|
975 |
-
if ('amount' in meds_features):
|
976 |
-
items_meds_amount = list(meds_features['amount'].keys())
|
977 |
-
values_meds_amount =[meds_features['amount'][i] for i in items_meds_amount ]
|
978 |
-
meds_amount = {"id" : items_meds_amount,
|
979 |
-
"value": values_meds_amount}
|
980 |
-
else:
|
981 |
-
meds_amount = {"id" : [],
|
982 |
-
"value": []}
|
983 |
-
|
984 |
-
meds = {"signal" : meds_sig,
|
985 |
-
"rate" : meds_rate,
|
986 |
-
"amount" : meds_amount}
|
987 |
-
|
988 |
-
|
989 |
-
yield int(hid), {
|
990 |
-
"label" : label,
|
991 |
-
"gender" : gender,
|
992 |
-
"ethnicity" : eth,
|
993 |
-
"insurance" : insurance,
|
994 |
-
"age" : age,
|
995 |
-
"COND" : cond_features,
|
996 |
-
"PROC" : procs,
|
997 |
-
"CHART/LAB" : charts,
|
998 |
-
"OUT" : outs,
|
999 |
-
"MEDS" : meds
|
1000 |
-
}
|
1001 |
-
|
1002 |
-
|
1003 |
-
|
1004 |
-
###########################################################ENCODED##################################################################
|
1005 |
-
|
1006 |
-
def _info_encoded(self):
|
1007 |
-
features = datasets.Features(
|
1008 |
-
{
|
1009 |
-
"label": datasets.ClassLabel(num_classes=2,names=["0", "1"]),
|
1010 |
-
"features" : datasets.Sequence(datasets.Value("float32")),
|
1011 |
-
}
|
1012 |
-
)
|
1013 |
-
return datasets.DatasetInfo(
|
1014 |
-
description=_DESCRIPTION,
|
1015 |
-
features=features,
|
1016 |
-
homepage=_HOMEPAGE,
|
1017 |
-
citation=_CITATION,
|
1018 |
-
)
|
1019 |
-
|
1020 |
-
def _generate_examples_encoded(self, filepath):
|
1021 |
-
path= './data/dict/'+self.config.name.replace(" ","_")+'/ethVocab'
|
1022 |
-
with open(path, 'rb') as fp:
|
1023 |
-
ethVocab = pickle.load(fp)
|
1024 |
-
|
1025 |
-
path= './data/dict/'+self.config.name.replace(" ","_")+'/insVocab'
|
1026 |
-
with open(path, 'rb') as fp:
|
1027 |
-
insVocab = pickle.load(fp)
|
1028 |
-
|
1029 |
-
genVocab = ['<PAD>', 'M', 'F']
|
1030 |
-
gen_encoder = LabelEncoder()
|
1031 |
-
eth_encoder = LabelEncoder()
|
1032 |
-
ins_encoder = LabelEncoder()
|
1033 |
-
gen_encoder.fit(genVocab)
|
1034 |
-
eth_encoder.fit(ethVocab)
|
1035 |
-
ins_encoder.fit(insVocab)
|
1036 |
-
with open(filepath, 'rb') as fp:
|
1037 |
-
dico = pickle.load(fp)
|
1038 |
-
|
1039 |
-
df = pd.DataFrame.from_dict(dico, orient='index')
|
1040 |
-
task=self.config.name.replace(" ","_")
|
1041 |
-
|
1042 |
-
for i, data in df.iterrows():
|
1043 |
-
concat_cols=[]
|
1044 |
-
dyn_df,cond_df,demo=concat_data(data,task,self.feat_cond,self.feat_proc,self.feat_out, self.feat_chart, self.feat_meds,self.feat_lab)
|
1045 |
-
dyn=dyn_df.copy()
|
1046 |
-
dyn.columns=dyn.columns.droplevel(0)
|
1047 |
-
cols=dyn.columns
|
1048 |
-
time=dyn.shape[0]
|
1049 |
-
for t in range(time):
|
1050 |
-
cols_t = [str(x) + "_"+str(t) for x in cols]
|
1051 |
-
concat_cols.extend(cols_t)
|
1052 |
-
demo['gender']=gen_encoder.transform(demo['gender'])
|
1053 |
-
demo['ethnicity']=eth_encoder.transform(demo['ethnicity'])
|
1054 |
-
demo['insurance']=ins_encoder.transform(demo['insurance'])
|
1055 |
-
label = data['label']
|
1056 |
-
demo=demo.drop(['label'],axis=1)
|
1057 |
-
X= getXY(dyn_df,cond_df,demo,concat_cols,self.concat)
|
1058 |
-
X=X.values.tolist()[0]
|
1059 |
-
yield int(i), {
|
1060 |
-
"label": label,
|
1061 |
-
"features": X,
|
1062 |
-
}
|
1063 |
-
######################################################DEEP###############################################################
|
1064 |
-
def _info_deep(self):
|
1065 |
-
features = datasets.Features(
|
1066 |
-
{
|
1067 |
-
"label": datasets.ClassLabel(num_classes=2,names=["0", "1"]),
|
1068 |
-
#"DEMO": datasets.Array2D(shape=(None, 4), dtype='int64') ,
|
1069 |
-
"DEMO": datasets.Sequence(datasets.Value("int64")),
|
1070 |
-
"COND" : datasets.Sequence(datasets.Value("int64")),
|
1071 |
-
#"COND" : datasets.Array2D(shape=(None, self.size_cond), dtype='int64') ,
|
1072 |
-
"MEDS" : datasets.Array2D(shape=(None, self.size_meds), dtype='int64') ,
|
1073 |
-
"PROC" : datasets.Array2D(shape=(None, self.size_proc), dtype='int64') ,
|
1074 |
-
"CHART/LAB" : datasets.Array2D(shape=(None, self.size_chart), dtype='int64') ,
|
1075 |
-
#"CHART/LAB" : datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
|
1076 |
-
"OUT" : datasets.Array2D(shape=(None, self.size_out), dtype='int64') ,
|
1077 |
-
|
1078 |
-
}
|
1079 |
-
)
|
1080 |
-
return datasets.DatasetInfo(
|
1081 |
-
description=_DESCRIPTION,
|
1082 |
-
features=features,
|
1083 |
-
homepage=_HOMEPAGE,
|
1084 |
-
citation=_CITATION,
|
1085 |
-
)
|
1086 |
-
|
1087 |
-
|
1088 |
-
def _generate_examples_deep(self, filepath):
|
1089 |
-
with open(filepath, 'rb') as fp:
|
1090 |
-
dico = pickle.load(fp)
|
1091 |
-
task=self.config.name.replace(" ","_")
|
1092 |
-
for key, data in dico.items():
|
1093 |
-
stat, demo, meds, chart, out, proc, lab, y = getXY_deep(data, task, self.feat_cond, self.feat_proc, self.feat_out, self.feat_chart, self.feat_meds,self.feat_lab)
|
1094 |
-
|
1095 |
-
verri=True
|
1096 |
-
if self.feat_proc:
|
1097 |
-
if (len(proc)<(self.timeW//self.bucket)):
|
1098 |
-
verri=False
|
1099 |
-
if self.feat_out:
|
1100 |
-
if (len(out)<(self.timeW//self.bucket)):
|
1101 |
-
verri=False
|
1102 |
-
if self.feat_chart:
|
1103 |
-
if (len(chart)<(self.timeW//self.bucket)):
|
1104 |
-
verri=False
|
1105 |
-
if self.feat_meds:
|
1106 |
-
if (len(meds)<(self.timeW//self.bucket)):
|
1107 |
-
verri=False
|
1108 |
-
if self.feat_lab:
|
1109 |
-
if (len(lab)<(self.timeW//self.bucket)):
|
1110 |
-
verri=False
|
1111 |
-
if verri:
|
1112 |
-
if self.data_icu:
|
1113 |
-
yield int(key), {
|
1114 |
-
'label': y,
|
1115 |
-
'DEMO': demo,
|
1116 |
-
'COND': stat,
|
1117 |
-
'MEDS': meds,
|
1118 |
-
'PROC': proc,
|
1119 |
-
'CHART/LAB': chart,
|
1120 |
-
'OUT': out,
|
1121 |
-
}
|
1122 |
-
else:
|
1123 |
-
yield int(key), {
|
1124 |
-
'label': y,
|
1125 |
-
'DEMO': demo,
|
1126 |
-
'COND': stat,
|
1127 |
-
'MEDS': meds,
|
1128 |
-
'PROC': proc,
|
1129 |
-
'CHART/LAB': lab,
|
1130 |
-
'OUT': out,
|
1131 |
-
}
|
1132 |
-
else:
|
1133 |
-
continue
|
1134 |
-
|
1135 |
-
|
1136 |
-
#############################################################################################################################
|
1137 |
-
def _info(self):
|
1138 |
-
self.path = self.create_cohort()
|
1139 |
-
self.size_cond, self.size_proc, self.size_meds, self.size_out, self.size_chart, self.size_lab, eth_vocab,gender_vocab,age_vocab,ins_vocab=vocab(self.config.name.replace(" ","_"),self.feat_cond,self.feat_proc,self.feat_out,self.feat_chart,self.feat_meds,self.feat_lab)
|
1140 |
-
|
1141 |
-
if self.encoding == 'concat' :
|
1142 |
-
return self._info_encoded()
|
1143 |
-
|
1144 |
-
elif self.encoding == 'aggreg' :
|
1145 |
-
return self._info_encoded()
|
1146 |
-
|
1147 |
-
elif self.encoding == 'tensor' :
|
1148 |
-
return self._info_deep()
|
1149 |
-
|
1150 |
-
else:
|
1151 |
-
return self._info_raw()
|
1152 |
-
|
1153 |
-
|
1154 |
-
def _split_generators(self, dl_manager):
|
1155 |
-
csv_dir = "./data/dict/"+self.config.name.replace(" ","_")
|
1156 |
-
if self.val_size > 0 :
|
1157 |
-
return [
|
1158 |
-
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": csv_dir+'/train_data.pkl'}),
|
1159 |
-
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": csv_dir+'/val_data.pkl'}),
|
1160 |
-
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": csv_dir+'/test_data.pkl'}),
|
1161 |
-
]
|
1162 |
-
else :
|
1163 |
-
return [
|
1164 |
-
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": csv_dir+'/train_data.pkl'}),
|
1165 |
-
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": csv_dir+'/test_data.pkl'}),
|
1166 |
-
]
|
1167 |
-
def _generate_examples(self, filepath):
|
1168 |
-
|
1169 |
-
if self.encoding == 'concat' :
|
1170 |
-
yield from self._generate_examples_encoded(filepath)
|
1171 |
-
|
1172 |
-
elif self.encoding == 'aggreg' :
|
1173 |
-
yield from self._generate_examples_encoded(filepath)
|
1174 |
-
|
1175 |
-
elif self.encoding == 'tensor' :
|
1176 |
-
yield from self._generate_examples_deep(filepath)
|
1177 |
-
else :
|
1178 |
-
yield from self._generate_examples_raw(filepath)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|