def convert_icd_9_to_10(s):
try:
return str(
_icd_mapping_9_to_10.get(s)
or _icd_mapping_9_to_10.get(icd9.search(s).parent.alt_code)
or _icd_mapping_9_to_10.get(icd9.search(s).parent.parent.alt_code))
except:
warnings.warn('Conversion failed: ' + str(s))
return s
def map_icd_hierarchy(s, version=9):
s = str(s)
code9 = icd9.search(s)
code10 = icd10.find(s)
if code9 is None and code10 is None:
raise Exception("Invalid ICD code", s)
if version == 9:
if code9 is not None:
return list(reversed([code9.alt_code] + code9.ancestors()))
else:
raise Exception("Invalid ICD version", s)
elif version == 10:
if code10 is not None:
return [code10.chapter, code10.block, code10.code[:3], code10.code]
else:
try:
# Attempt to convert from version 9 to 10
s_ = convert_icd_9_to_10(code9.alt_code)
code = icd10.find(s_)
if code is None:
# Fall back to version 9
return list(reversed([code9.alt_code] + code9.ancestors()))
else:
return [code.chapter, code.block, code.code[:3], code.code]
except:
warnings.warn('Conversion failed: ' + str(s))
return list(reversed([code9.alt_code] + code9.ancestors()))
# raise Exception('Conversion error: ' + str(s))
else:
raise Exception("Invalid ICD version", s)
def get_comorbidites_x(self, hadm_id):
comorbidities = self.comorbidites[hadm_id]
comorbidities_x = [0, 0]
for comorbidity_code in comorbidities:
info = icd.search(comorbidity_code)
while info is not None:
description = info.short_desc
if description == 'Infectious And Parasitic Diseases':
comorbidities_x[0] = 1
break
elif description == 'Diseases Of The Skin And Subcutaneous Tissue':
comorbidities_x[1] = 1
break
else:
info = info.parent
return comorbidities_x
def ccs_kg_building(graph_file, out_file):
_TEST_RATIO = 0.2
_VALIDATION_RATIO = 0.1
entities = {}
code2desc = {}
relation = {
'is_parent_of': 0,
'is_child_of': 1
}
train_set = set()
infd = open(graph_file, 'r')
_ = infd.readline()
# build dictionary of entities
for line in infd:
tokens = line.strip().split(',')
icd9 = tokens[0][1:-1].strip()
cat1 = tokens[1][1:-1].strip()
desc1 = 'A_L1_' + cat1
cat2 = tokens[3][1:-1].strip()
desc2 = 'A_L2_' + cat2
cat3 = tokens[5][1:-1].strip()
desc3 = 'A_L3_' + cat3
cat4 = tokens[7][1:-1].strip()
desc4 = 'A_L4_' + cat4
icd9 = 'D_' + icd9
if icd9 not in entities:
entities[icd9] = len(entities)
if desc1 not in entities:
entities[desc1] = len(entities)
if len(cat2) > 0:
if desc2 not in entities:
entities[desc2] = len(entities)
if len(cat3) > 0:
if desc3 not in entities:
entities[desc3] = len(entities)
if len(cat4) > 0:
if desc4 not in entities:
entities[desc4] = len(entities)
sr = search(icd9[2:])
i = 1
while sr is None:
# print(sr)
l = len(icd9)
if l-i == 0:
break
sr = search(icd9[2:l-i])
i += 1
if sr is not None:
ds = str(sr).split(':')
if ds[2] == 'None':
code2desc[icd9] = ds[1]
else:
code2desc[icd9] = ds[2]
infd.close()
# add root_code
entities['A_ROOT'] = len(entities)
rootCode = entities['A_ROOT']
infd = open(infile, 'r')
infd.readline()
for line in infd:
tokens = line.strip().split(',')
icd9 = tokens[0][1:-1].strip()
cat1 = tokens[1][1:-1].strip()
desc1 = 'A_L1_' + cat1
cat2 = tokens[3][1:-1].strip()
desc2 = 'A_L2_' + cat2
cat3 = tokens[5][1:-1].strip()
desc3 = 'A_L3_' + cat3
cat4 = tokens[7][1:-1].strip()
desc4 = 'A_L4_' + cat4
icd9 = 'D_' + icd9
# if icd9 not in entities:
# continue
icdCode = entities[icd9]
if len(cat4) > 0:
code4 = entities[desc4]
code3 = entities[desc3]
code2 = entities[desc2]
code1 = entities[desc1]
train_set.add(str(rootCode) + ',' + str(code1) + ',0')
train_set.add(str(code1) + ',' + str(rootCode) + ',1')
train_set.add(str(code1) + ',' + str(code2) + ',0')
train_set.add(str(code2) + ',' + str(code1) + ',1')
train_set.add(str(code2) + ',' + str(code3) + ',0')
train_set.add(str(code3) + ',' + str(code2) + ',1')
train_set.add(str(code3) + ',' + str(code4) + ',0')
train_set.add(str(code4) + ',' + str(code3) + ',1')
train_set.add(str(code4) + ',' + str(icdCode) + ',0')
train_set.add(str(icdCode) + ',' + str(code4) + ',1')
elif len(cat3) > 0:
code3 = entities[desc3]
code2 = entities[desc2]
code1 = entities[desc1]
train_set.add(str(rootCode) + ',' + str(code1) + ',0')
train_set.add(str(code1) + ',' + str(rootCode) + ',1')
train_set.add(str(code1) + ',' + str(code2) + ',0')
train_set.add(str(code2) + ',' + str(code1) + ',1')
train_set.add(str(code2) + ',' + str(code3) + ',0')
train_set.add(str(code3) + ',' + str(code2) + ',1')
train_set.add(str(code3) + ',' + str(icdCode) + ',0')
train_set.add(str(icdCode) + ',' + str(code3) + ',1')
elif len(cat2) > 0:
code2 = entities[desc2]
code1 = entities[desc1]
train_set.add(str(rootCode) + ',' + str(code1) + ',0')
train_set.add(str(code1) + ',' + str(rootCode) + ',1')
train_set.add(str(code1) + ',' + str(code2) + ',0')
train_set.add(str(code2) + ',' + str(code1) + ',1')
train_set.add(str(code2) + ',' + str(icdCode) + ',0')
train_set.add(str(icdCode) + ',' + str(code2) + ',1')
else:
code1 = entities[desc1]
train_set.add(str(rootCode) + ',' + str(code1) + ',0')
train_set.add(str(code1) + ',' + str(rootCode) + ',1')
train_set.add(str(code1) + ',' + str(icdCode)+ ',0')
train_set.add(str(icdCode) + ',' + str(code1) + ',1')
outfd = open(out_file + 'relation2id', 'w')
for k, v in relation.items():
outfd.write(str(v) + ',' + k + '\n')
outfd.close()
outfd = open(out_file + 'entity2id', 'w')
for k, v in entities.items():
outfd.write(str(v) + ',' + k + '\n')
outfd.close()
edges_ls = list(train_set)
data_size = len(edges_ls)
ind = np.random.permutation(data_size)
nTest = int(_TEST_RATIO * data_size)
nValid = int(_VALIDATION_RATIO * data_size)
test_indices = ind[:nTest]
valid_indices = ind[nTest:nTest + nValid]
train_indices = ind[nTest + nValid:]
train = [edges_ls[i] for i in train_indices]
valid = [edges_ls[i] for i in valid_indices]
test = [edges_ls[i] for i in test_indices]
outfd = open(out_file + 'train_file', 'w')
for edge in train:
outfd.write(edge + '\n')
outfd.close()
outfd = open(out_file + 'valid_file', 'w')
for edge in valid:
outfd.write(edge + '\n')
outfd.close()
outfd = open(out_file + 'test_file', 'w')
for edge in test:
outfd.write(edge + '\n')
outfd.close()
with open(out_file + 'code2desc.pickle', 'wb') as handle:
pickle.dump(code2desc, handle, protocol=pickle.HIGHEST_PROTOCOL)
print(len(code2desc))