/
compute_ndfrt_baseline.py
122 lines (106 loc) · 4.14 KB
/
compute_ndfrt_baseline.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
from collections import Counter
from collections import defaultdict
import csv
import pickle
import numpy as np
from tqdm import tqdm
import pandas as pd
def flatten(ll):
return [item for sublist in ll for item in sublist]
def is_match(code, code_set):
for i in range(1, len(code)+1):
if code[:i] in code_set:
return True
return False
def overlap(code_set_1, code_set_2):
for c in code_set_1:
if is_match(c, code_set_2):
return True
for c in code_set_2:
if is_match(c, code_set_1):
return True
return False
problem_snomed, problem_icd10, problem_icd9 = defaultdict(list), defaultdict(list), defaultdict(list)
print("loading problem definitions")
with open('data/problem_codes_all.csv') as f:
r = csv.reader(f)
header = next(r)
for row in r:
if row[1] == 'icd10':
problem_icd10[row[0]].append(row[2])
if row[1] == 'icd9':
problem_icd9[row[0]].append(row[2])
if row[1] == 'snomed':
problem_snomed[row[0]].append(row[2])
df = pd.read_csv('data/test_probs.csv')
adf = pd.read_csv('data/all.csv', names=['problem', 'relationType', 'target', 'label'])
meds = sorted([med.split('_')[1] for med in adf[adf['relationType'] == 'medication']['target'].unique()])
print(f"num all meds: {len(set(meds))}")
print(f"num rxnorm meds: {len(set([med for med in meds if 'XXXXX' not in med]))}")
med_may_treat = {'snomed': defaultdict(list), 'icd9': defaultdict(list), 'icd10': defaultdict(list)}
with open('data/med_may_treat.csv') as f:
r = csv.reader(f)
for row in r:
med, dx, dx_type = row
med_may_treat[dx_type][med].append(dx)
print("processing NEGATIVES")
num_negs = Counter()
num_pos_negs = Counter()
for ix, triple in enumerate(df.itertuples()):
if triple.label == 1:
if triple.relationType != 'medication':
continue
target = triple.target.split('_')[1]
m_sno = flatten(med_may_treat['snomed'][target])
m_icd9 = flatten(med_may_treat['icd9'][target])
m_icd10 = flatten(med_may_treat['icd10'][target])
p_sno = problem_snomed[triple.problem]
p_icd9 = problem_icd9[triple.problem]
p_icd10 = problem_icd10[triple.problem]
num_negs[(triple.problem, triple.relationType)] += 1
if len(set(m_sno).intersection(set(p_sno))) > 0:
num_pos_negs[(triple.problem, triple.relationType)] += 1
elif overlap(m_icd10, p_icd10):
num_pos_negs[(triple.problem, triple.relationType)] += 1
elif overlap(m_icd9, p_icd9):
num_pos_negs[(triple.problem, triple.relationType)] += 1
print("\n\nprocessing POSITIVES")
ranks = []
matches = 0
def rx_prob_match(target, problem):
p_sno = problem_snomed[problem]
p_icd9 = problem_icd9[problem]
p_icd10 = problem_icd10[problem]
m_sno = med_may_treat['snomed'][target]
m_icd9 = med_may_treat['icd9'][target]
m_icd10 = med_may_treat['icd10'][target]
if len(set(m_sno).intersection(set(p_sno))) > 0:
return 'snomed'
elif overlap(m_icd10, p_icd10):
return 'icd10'
elif overlap(m_icd9, p_icd9):
return 'icd9'
return None
for triple in df.itertuples():
if triple.label != 2 or triple.relationType != 'medication':
continue
target = triple.target.split('_')[1]
nn = num_negs[(triple.problem, triple.relationType)]
npn = num_pos_negs[(triple.problem, triple.relationType)]
route = rx_prob_match(target, triple.problem)
if route != None:
rank = npn / 2 + 1
matches += 1
else:
# no match
rank = np.median(np.arange(npn+1, nn+1+1))
probs = [prob for prob in problem_icd10 if rx_prob_match(target, prob) is not None]
ranks.append(rank)
ranks = np.array(ranks)
mr = np.mean(ranks)
mrr = np.mean(1./ranks)
hits_at_1 = np.mean([rank <= 1 for rank in ranks])
hits_at_5 = np.mean([rank <= 5 for rank in ranks])
hits_at_10 = np.mean([rank <= 10 for rank in ranks])
hits_at_30 = np.mean([rank <= 30 for rank in ranks])
print(f"MR: {mr}, MRR: {mrr}, H@1: {hits_at_1}, H@5: {hits_at_5}, H@10: {hits_at_10}, H@30: {hits_at_30}")