def get_order_type(dataset_id, ordertype):
data = dt.get(dataset_id).aslist()
#pat_ids = [int(d['PAT_ID'] for d in data])
#if pat_ids == PAT_ID:
filtered_data = [x for x in data if x['ORDER_CATALOG_TYPE'] == ordertype ]
filtered_data.sort(key=lambda x: to_data_time(x['ORDER_DTM']), reverse=True)
return filtered_data
def calc_diff(id1, id2, direction, ops):
"""
Calculate a detailed difference between two tables
:param id1:
:param id2:
:param direction:
:param ops:
:return:
"""
ds1 = dataset.get(id1)
ds2 = dataset.get(id2)
def stratify_matrix(m):
# NOTE: ids must match with the hard-coded ones in taco/src/data_set_selector.ts -> prepareTCGAData()
row_strat = dataset.get(m.id + '4CnmfClustering')
col_strat = dataset.get(m.id + 'TreeClusterer1')
data = m.asnumpy()
rows = m.rows() if row_strat is None else row_strat.rows()
cols = m.cols() if col_strat is None else col_strat.rows()
if row_strat is not None:
rowids = list(m.rowids())
row_indices = [rowids.index(o) for o in row_strat.rowids()]
data = data[row_indices, ...]
if col_strat is not None:
colids = list(m.colids())
col_indices = [colids.index(o) for o in col_strat.rowids()]
data = data[..., col_indices]
return Table(rows, cols, data)
# create the table object
table1 = stratify_matrix(ds1)
table2 = stratify_matrix(ds2)
diff_finder = DiffFinder(table1, table2, ds1.rowtype, ds2.coltype,
direction)
t2 = timeit.default_timer()
d = diff_finder.generate_diff(ops)
t3 = timeit.default_timer()
_log.debug("TIMER: time to generate diff", t3 - t2)
if isinstance(d, Diff):
d.add_union(diff_finder.union)
return d
def get_info(ids, dataset_id):
data = dt.get(dataset_id).aslist()
# find the first entry for each patient
result = {}
for id in ids:
result[id] = []
for row in data:
if int(row['PAT_ID']) in ids:
result[int(row['PAT_ID'])].append(row)
return result
def stratify_matrix(m):
# NOTE: ids must match with the hard-coded ones in taco/src/data_set_selector.ts -> prepareTCGAData()
row_strat = dataset.get(m.id + '4CnmfClustering')
col_strat = dataset.get(m.id + 'TreeClusterer1')
data = m.asnumpy()
rows = m.rows() if row_strat is None else row_strat.rows()
cols = m.cols() if col_strat is None else col_strat.rows()
if row_strat is not None:
rowids = list(m.rowids())
row_indices = [rowids.index(o) for o in row_strat.rowids()]
data = data[row_indices, ...]
if col_strat is not None:
colids = list(m.colids())
col_indices = [colids.index(o) for o in col_strat.rowids()]
data = data[..., col_indices]
return Table(rows, cols, data)
def get_first_info(PAT_ID):
data = dt.get('Orders').aslist()
# find the first entry for each patient
first_ent = []
first_time = pd.datetime.today()
for row in data:
if row['PAT_ID'] == PAT_ID:
if to_data_time(row['ORDER_DTM']) < first_time:
first_ent = row
first_time = to_data_time(row['ORDER_DTM'])
return first_ent
def get_first_info(PAT_ID):
data = dt.get('PRO').aslist()
# find the first entry for each patient
first_ent = []
first_time = pd.datetime.today()
for row in data:
if row['PAT_ID'] == PAT_ID:
if to_date_time(row['ASSESSMENT_START_DTM']) < first_time:
first_ent = row
first_time = to_date_time(row['ASSESSMENT_START_DTM'])
return first_ent
def get_latest_info():
data = dt.get('Demo').aslist()
pat_ids = set([int(d['PAT_ID']) for d in data])
pats = []
pat_weights = []
for id in pat_ids:
temp_pat = []
for row in data:
if row['PAT_ID'] == id:
temp_pat.append(row)
temp_pat.sort(key=lambda x: to_date_time(x['ADM_DATE']), reverse=True)
pats.append(temp_pat[0])
pat_weights.append(
[t['WEIGHT_KG'] for t in temp_pat if t['WEIGHT_KG']])
return jsonify({'rows': pats, 'WEIGHT_KG': pat_weights})
def get_similarity_score(PAT_ID, dataset_id):
data = dt.get(dataset_id).aslist()
pat_ids = set([int(d['PAT_ID']) for d in data])
# find the first entry for each patient
pats = {}
#for id in pat_ids:
# pats[id] = get_first_info(id)
for row in data:
pats[int(row['PAT_ID'])] = row
scores = {}
target_pat = pats[PAT_ID]
for id in pat_ids:
similarity_score = 0
if id == PAT_ID: ## NOTE!
continue
curr_pat = pats[id]
for h in list(weights_Demo.keys()):
if weights_Demo[h] == 0:
continue
elif not curr_pat[h] or not target_pat[h]:
similarity_score += 0.5 * weights_Demo[h]
elif h in categorical_data:
similarity_score += weights_Demo[h] if target_pat[
h] == curr_pat[h] else 0
elif h in numerical_data:
similarity_score += weights_Demo[h] if target_pat[
h] == curr_pat[h] else 0 ## within range
elif h in years: ## age! ## within range
similarity_score += weights_Demo[h] if float(
extract_year(target_pat[h])) == float(
extract_year(curr_pat[h])) else 0
else:
similarity_score += weights_Demo[h] if target_pat[
h] == curr_pat[h] else 0
scores[id] = similarity_score
return scores
def get_info_by_functions(id):
my_data = dt.get(id)
#range = rng.RangeElem(2, 5, 1)
return jsonify({
'id': id,
'my_data': my_data,
'rows': my_data.rows(),
'data': my_data.asjson(),
'rowids': my_data.rowids(),
'aslist': my_data.aslist(),
'aspandas': my_data.aspandas(),
'columns': my_data.columns,
'CSVEntry.to_description': my_data.to_description(),
'CSVEntry.idtypes': my_data.idtypes() #,
#'asList( range(2,5,1) )': my_data.aslist(range),
#'rows( range(2,5,1) )': my_data.rows(range),
#'aspandas( range(2,5,1) )': my_data.aspandas(range)
})
def get_all_info_for_pat(dataset_id, PAT_ID):
my_data = dt.get(dataset_id)
info_rows = []
indices = []
for i in range(0, len(my_data.rows())):
if int(my_data.aspandas()['PAT_ID'][my_data.rows()[i]]) == int(PAT_ID):
indices.append(i)
info_rows.append(my_data.aslist()[i])
if dataset_id == 'Demo':
info_rows.sort(key=lambda r: to_date_time(r["ADM_DATE"]))
elif dataset_id == 'PRO':
info_rows.sort(key=lambda r: to_date_time(r["ASSESSMENT_START_DTM"]))
elif dataset_id == 'PT':
info_rows.sort(key=lambda r: to_date_time(r["ADM_DATE"]))
elif dataset_id == 'VAS':
info_rows.sort(key=lambda r: to_date_time(r["RECORDED_TIME"]))
return info_rows
def get_stat(dataset):
my_data = dt.get(dataset_hash['Demo'][dataset])
data = my_data.aslist()
length = 0
gender = [0, 0]
bmi = [0, 0, 0, 0, 0, 0, 0]
age = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
today = pd.datetime.today()
for row in data:
length += 1
if row['PAT_GENDER'] == 'F':
gender[0] += 1
elif row['PAT_GENDER'] == 'M':
gender[1] += 1
if not row['BMI']:
bmi[0] += 1
elif row['BMI'] <= 18:
bmi[1] += 1
elif row['BMI'] > 30:
bmi[6] += 1
else:
bmi[int(math.ceil((row['BMI'] - 15) / 3))] += 1
pat_age = int(
(today - to_date_time(row['PAT_BIRTHDATE'])).days / 365.25)
if pat_age <= 10:
age[0] += 1
elif pat_age > 100:
age[10] += 1
else:
age[int(math.floor((pat_age - 1) / 10))] += 1
return jsonify({
'length': length,
'GENDER': gender,
'BMI': bmi,
'AGE': age
})
def get_data(id):
return jsonify({'data': dt.get(id)})
def get_similarity_score(PAT_ID, dataset_id):
data = dt.get(dataset_id).aslist()
# divide data to 3 groups
# find the first entry for each patient in each group
pat_ids = set([int(d['PAT_ID']) for d in data])
groups = {
'PROMIS Bank v1.2 - Physical Function': {},
'Oswestry Index (ODI)': {},
'PROMIS Bank v1.0 - Depression': {}
}
for row in data:
promisType = row['FORM']
if int(row['PAT_ID']) in groups[promisType]:
if to_date_time(groups[promisType][row['PAT_ID']]
['ASSESSMENT_START_DTM']) > to_date_time(
row['ASSESSMENT_START_DTM']):
groups[promisType][int(row['PAT_ID'])] = row
else:
groups[promisType][int(row['PAT_ID'])] = row
physical = groups['PROMIS Bank v1.2 - Physical Function'][int(PAT_ID)]
oswerstry = groups['Oswestry Index (ODI)'][int(PAT_ID)]
depression = groups['PROMIS Bank v1.0 - Depression'][int(PAT_ID)]
scores = {}
for id in pat_ids:
if id == int(PAT_ID):
continue
# only promis score
similarity_score = 0
if physical and id in groups['PROMIS Bank v1.2 - Physical Function']:
similarity_score += weights_PRO['SCORE'] if abs(
int(physical['SCORE']) -
int(groups['PROMIS Bank v1.2 - Physical Function'][id]
['SCORE'])) <= 10 else 0 ## within range of 10
else:
similarity_score += weights_PRO['SCORE'] * 0.5
if oswerstry and id in groups['Oswestry Index (ODI)']:
similarity_score += weights_PRO['SCORE'] if abs(
int(oswerstry['SCORE']) -
int(groups['Oswestry Index (ODI)'][id]['SCORE'])
) <= 10 else 0 ## within range of 10
else:
similarity_score += weights_PRO['SCORE'] * 0.5
if depression and id in groups['PROMIS Bank v1.0 - Depression']:
similarity_score += weights_PRO['SCORE'] if abs(
int(depression['SCORE']) -
int(groups['PROMIS Bank v1.0 - Depression'][id]['SCORE'])
) <= 10 else 0 ## within range of 10
else:
similarity_score += weights_PRO['SCORE'] * 0.5
scores[id] = similarity_score
return scores
def get_col_by_name(id, col_name):
my_data = dt.get(id)
cols = my_data.aspandas()
return jsonify({'cols': cols[col_name]})
def get_col_titles(id):
my_data = dt.get(id)
return jsonify({'cols': my_data.columns})
def get_filteres_orders(dataset_id, order):
data = dt.get(dataset_id).aslist()
filtered_data = [x for x in data if x['PRIMARY_MNEMONIC'].upper() == order ]
filtered_data.sort(key=lambda x: to_data_time(x['ORDER_DTM']), reverse=True)
return filtered_data
def get_all_rows(dataset_id):
my_data = dt.get(dataset_id)
rows = my_data.aslist()
return jsonify({dataset_id: rows[:50]})
def get_row_by_index(id, index):
my_data = dt.get(id)
rows = my_data.aslist()
return jsonify({'row': rows[int(index)]})