def determine_iud_expulsion(document: Document):
if document.has_patterns(IUD):
for section in document.select_sentences_with_patterns(IUD):
history = bool(section.has_patterns(PREVIOUS))
if section.has_patterns(INCORRECT, PLACEMENT, has_all=True) or \
section.has_patterns(MALPOSITION) or \
section.has_patterns(DISPLACEMENT):
if not section.has_patterns(IN_CERVIX):
yield ExpulsionStatus.MALPOSITION, history, section.text
if section.has_patterns(LOWER_UTERINE_SEGMENT):
yield ExpulsionStatus.LOWER_UTERINE_SEGMENT, history, section.text
if section.has_patterns(COMPLETE):
yield ExpulsionStatus.EXPULSION, history, section.text
if section.has_patterns(PARTIAL_EXP):
yield ExpulsionStatus.PARTIAL, history, section.text
if section.has_patterns(MISSING, STRINGS, has_all=True):
yield ExpulsionStatus.MISSING_STRING, history, section.text
if section.has_patterns(PROPER_LOCATION):
yield ExpulsionStatus.PROPER_PLACEMENT, history, section.text
if section.has_patterns(IN_UTERUS):
yield ExpulsionStatus.IN_UTERUS, history, section.text
sects = document.select_all_sentences_with_patterns(STRING_HANGING)
if sects:
yield ExpulsionStatus.HANGING_STRING, None, sects.text
else:
yield ExpulsionStatus.SKIP, None, document.text
def determine_iud_brand(document: Document):
if document.has_patterns(EXCLUDE):
return [(BrandStatus.SKIP, False, None)]
if document.has_patterns(*IUD, ignore_negation=True):
brands = []
for section in document.select_sentences_with_patterns(
*IUD, neighboring_sentences=1):
# scheduled = bool(section.has_patterns(SCHEDULED))
if section.has_patterns(PARAGARD, COPPER):
brands.append((BrandStatus.PARAGARD,
section.has_patterns(USING), section.text))
if section.has_patterns(MIRENA):
brands.append((BrandStatus.MIRENA, section.has_patterns(USING),
section.text))
if section.has_patterns(LILETTA):
brands.append((BrandStatus.LILETTA,
section.has_patterns(USING), section.text))
if section.has_patterns(KYLEENA):
brands.append((BrandStatus.KYLEENA,
section.has_patterns(USING), section.text))
if section.has_patterns(SKYLA):
brands.append((BrandStatus.SKYLA, section.has_patterns(USING),
section.text))
if not brands: # only look for LNG if no brand name
if document.has_patterns(LNG):
brands.append((BrandStatus.LNG, False, document.text))
else:
brands.append((BrandStatus.NONE, False, None))
return brands
else:
return [(BrandStatus.SKIP, False, None)]
def determine_iud_removal(document: Document):
if document.has_patterns(*ALL, ignore_negation=True):
section_text = []
for section in document.select_sentences_with_patterns(IUD):
if section.has_pattern(REMOVE_BY):
continue
# these definitely have correct language
if section.has_patterns(*ALL):
# require either REMOVE/PLAN since this could have other refs
if section.has_patterns(DEF_REMOVE):
yield RemoveStatus.DEF_REMOVE, section.text
if section.has_patterns(DEF_REPLACE):
yield RemoveStatus.DEF_REPLACE, section.text
if section.has_patterns(PROB_REMOVE):
yield RemoveStatus.REMOVE, section.text
if section.has_patterns(TOOL):
yield RemoveStatus.TOOL_REMOVE, section.text
if section.has_patterns(REMOVE):
yield RemoveStatus.REMOVE, section.text
if section.has_patterns(PLAN):
yield RemoveStatus.PLAN, section.text
else:
section_text.append(section.text)
if section_text:
yield RemoveStatus.NONE, ' '.join(section_text)
else:
yield RemoveStatus.SKIP, document.text
else:
yield RemoveStatus.SKIP, document.text
def determine_iud_expulsion_rad(document: Document):
sections = document.split(SECTIONS)
start_section = sections.get_sections('HST', 'SAS', 'HISTORY',
'CLINICAL INFORMATION',
'CLINICAL HISTORY AND QUESTION')
impression = sections.get_sections('IMPRESSION', 'IMPRESSIONS', 'IMP')
other = sections.get_sections('FINDINGS', 'TRANSVAGINAL', 'FINDING')
found = False
if impression:
if impression.has_patterns(PARTIAL_EXP):
found = True
yield ExpulsionStatus.PARTIAL, impression.text
if impression.has_patterns(NOT_SEEN_IUD, IUD_NOT_SEEN):
found = True
yield ExpulsionStatus.LOST, impression.text
if impression.has_patterns(MALPOSITION, MALPOSITION_IUD):
found = True
yield ExpulsionStatus.MALPOSITION, impression.text
if impression.has_patterns(PROPER_LOCATION):
found = True
yield ExpulsionStatus.PROPER_PLACEMENT, impression.text
if impression.has_patterns(IN_UTERUS):
found = False # don't count this one
yield ExpulsionStatus.IN_UTERUS, impression.text
if impression.has_patterns(LOWER_UTERINE_SEGMENT):
found = False
yield ExpulsionStatus.LOWER_UTERINE_SEGMENT, impression.text
if found:
return
elif start_section.has_patterns(IUD, IUD_PRESENT, has_all=True) or \
start_section.has_patterns(STRING, VISIBLE, has_all=True):
if other.has_patterns(PARTIAL_EXP):
yield ExpulsionStatus.PARTIAL, other.text
if other.has_patterns(MALPOSITION, MALPOSITION_IUD):
yield ExpulsionStatus.MALPOSITION, other.text
if other.has_patterns(PROPER_LOCATION):
yield ExpulsionStatus.PROPER_PLACEMENT, other.text
if other.has_patterns(IN_UTERUS):
yield ExpulsionStatus.IN_UTERUS, other.text
if other.has_patterns(LOWER_UTERINE_SEGMENT):
yield ExpulsionStatus.LOWER_UTERINE_SEGMENT, other.text
else:
sentences = list(document.select_sentences_with_patterns(IUD))
if sentences:
for sentence in sentences:
if sentence.has_patterns(PARTIAL_EXP):
yield ExpulsionStatus.PARTIAL, sentence.text
if sentence.has_patterns(MALPOSITION, MALPOSITION_IUD):
yield ExpulsionStatus.MALPOSITION, sentence.text
if sentence.has_patterns(PROPER_LOCATION):
yield ExpulsionStatus.PROPER_PLACEMENT, sentence.text
if sentence.has_patterns(IN_UTERUS):
yield ExpulsionStatus.IN_UTERUS, sentence.text
if sentence.has_patterns(LOWER_UTERINE_SEGMENT):
yield ExpulsionStatus.LOWER_UTERINE_SEGMENT, sentence.text
else:
yield ExpulsionStatus.SKIP, None
def determine_parity(document: Document):
res = document.get_patterns(GRAVIDA_0, PARA_0, NULLIPAROUS, CHILD_0,
names=[ParitySource.GRAVIDA, ParitySource.PARITY,
ParitySource.PARITY, ParitySource.CHILDREN])
if res:
text, src = res
if text:
return ParityStatus.P0, text, src
# patterns for grav and para; confirm grav >= para
res = document.get_patterns(G_PARA_N, index=(0, 1, 2))
if res:
text, capt_g, capt_p = res
try:
g = int(capt_g)
p = int(capt_p)
except ValueError:
raise ValueError(f'Values of g/p do not appear to be numeric: {capt_p}, {capt_g} in {text}')
except TypeError:
raise TypeError(f'Values of g/p are None (or similar): {capt_p}, {capt_g} in {text}')
else:
if len(capt_p) == 4:
capt_p = capt_p[0]
p = int(capt_p)
if g >= p:
status = extract_status(capt_p)
if status:
return status, text, ParitySource.PARITY
elif g == p - 1: # common error: counting twins as extra parity
logging.warning(f'Including Gravida {g}, Parity {p} '
f'(assuming twins counted as extra parity): {document.name}')
status = extract_status(capt_g) # use gravida not parity
if status:
return status, text, ParitySource.GRAVIDA
else:
logging.info(f'Gravida {g} < Parity {p}: {document.name}')
res = document.get_patterns(PARA_NNNN, PARA_N, CHILD_N, CHILD_NUM, index=(0, 1),
names=[ParitySource.PARITY, ParitySource.PARITY,
ParitySource.CHILDREN, ParitySource.CHILDREN])
if res and res[0]:
(text, captured), src = res
if captured:
status = extract_status(captured)
if status:
return status, text, src
else:
# e.g., "para 13 %", "multipara @ 32+2/6weeks
logging.info(f'Unrecognized parity value for {document.name}: {status} in "{text}" from {src}')
text = document.get_patterns(MULTIPAROUS)
if text:
return ParityStatus.MULTIPAROUS, text, ParitySource.PARITY
return ParityStatus.SKIP, None, ParitySource.NONE
def determine_difficult_insertion(document: Document):
"""
:param document:
:return:
"""
if document.has_patterns(NOT_IUD_INSERTION):
yield DiffInsStatus.SKIP, None
elif document.has_patterns(IUD, ignore_negation=True):
sent = document.get_pattern(SUCCESSFUL_INSERTION)
if sent:
yield DiffInsStatus.SUCCESSFUL, sent
sent = document.get_pattern(CANNOT_PLACE)
if sent:
yield DiffInsStatus.UNSUCCESSFUL, sent
if document.has_patterns(INSERTION, ignore_negation=True):
for section in document.select_sentences_with_patterns(INSERTION, neighboring_sentences=1):
if section.has_patterns(UNSUCCESSFUL_INSERTION, has_all=True):
yield DiffInsStatus.UNSUCCESSFUL, section.text
if section.has_patterns(EASY_INSERTION):
yield DiffInsStatus.NOT_DIFFICULT, section.text
continue
if section.has_patterns(PROVIDER):
yield DiffInsStatus.PROVIDER_STATEMENT, section.text
if section.has_patterns(US, US_USED, has_all=True):
yield DiffInsStatus.ULTRASOUND_GUIDANCE, section.text
if section.has_patterns(PARACERV):
yield DiffInsStatus.PARACERVICAL_BLOCK, section.text
if section.has_patterns(MISOPROSTOL):
yield DiffInsStatus.MISOPROSTOL, section.text
if section.has_patterns(CERV_DIL):
yield DiffInsStatus.CERVICAL_DILATION, section.text
def determine_iud_insertion(document: Document):
# discusses iud
if document.has_patterns(IUD, INSERTION, has_all=True):
section = document.select_all_sentences_with_patterns(IUD, INSERTION, STRINGS,
negation=[HISTORICAL, APPOINTMENT, DATE, NEGATED],
neighboring_sentences=1)
if section:
if section.has_patterns(PRE_SUCCESS):
return InsertionStatus.SUCCESS, section.text
elif section.has_patterns(UNSUCCESSFUL, NOT_SUCCESSFUL):
return InsertionStatus.FAILED, section.text
elif section.has_patterns(HYPOTHETICAL):
return InsertionStatus.HYPOTHETICAL, section.text
elif section.has_patterns(POST_SUCCESS):
return InsertionStatus.SUCCESS, section.text
elif section.has_patterns(POST_OP):
return InsertionStatus.LIKELY_SUCCESS, section.text
else:
return InsertionStatus.UNKNOWN, section.text
else:
return InsertionStatus.UNKNOWN, None
return InsertionStatus.NO_MENTION, None
def determine_iud_perforation(document: Document):
if document.has_patterns(*ALL, ignore_negation=True):
# see if any sentences that contain "IUD" also contain perf/embedded
for section in document.select_sentences_with_patterns(IUD):
date = section.get_pattern(DATE_PAT)
if section.has_patterns(COMPLETE):
yield PerforationStatus.COMPLETE, section.text, date
elif section.has_patterns(PARTIAL):
yield PerforationStatus.PARTIAL, section.text, date
elif section.has_patterns(PERFORATION):
if section.has_pattern(POSSIBLE):
yield PerforationStatus.POSSIBLE, section.text, date
else:
yield PerforationStatus.PERFORATION, section.text, date
# elif section.has_patterns(EMBEDDED):
# yield PerforationStatus.EMBEDDED, section.text, date
# elif section.has_patterns(MIGRATED):
# yield PerforationStatus.MIGRATED, section.text, date
# check for laparoscopic removal -> suggests complete perf
if section.has_patterns(LAPAROSCOPIC_REMOVAL):
yield PerforationStatus.LAPAROSCOPIC_REMOVAL, section.text, date
yield PerforationStatus.NONE, document.text, None
else:
yield PerforationStatus.SKIP, None, None
def get_next_from_corpus(directory=None,
directories=None,
version=None,
connections=None,
skipper=None,
start=0,
end=None,
filenames=None,
encoding='utf8'):
"""
:param filenames:
:param encoding:
:param connections:
:param directories: list of directories to look through
:param skipper:
:param directory: first to look through (for backwards compatibility)
:param version: text|lemma|token
:param start:
:param end:
:return: iterator yielding documents
"""
i = -1
for doc_name, path, text in itertools.chain(
get_next_from_directory(directory, directories, version, filenames,
encoding),
get_next_from_connections(*connections or list())):
if skipper and doc_name in skipper:
continue
i += 1
if i < start:
continue
elif end and i >= end:
break
if not text and not path: # one of these required
continue
yield Document(doc_name, file=path, text=text)
def determine_breastfeeding(document: Document, expected=None):
my_result = partial(Result, expected=expected)
has_boilerplate = False
document = document.remove_patterns(BF_BOILERPLATE_SECTION)
if not document: # only boilerplate
return
if document.has_patterns(BF_BOILERPLATE_EXCLUDE, BF_UNKNOWN):
return
if document.has_patterns(BF_BOILERPLATE):
yield my_result(BreastfeedingStatus.BOILERPLATE)
has_boilerplate = True
non_exact_count = 0
non_exact_count_snippets = []
for sentence in document.sentences:
# look for presence of "nutrition:" or "feeding:" followed by not info about breast
if matches_nutrition_not_bf(sentence.text):
yield my_result(BreastfeedingStatus.OTHER_NUTRITION, text=sentence.text)
for section in document.select_sentences_with_patterns(ANY_BREAST):
found_bf = False
# pre boilerplate patterns: exact/not confused with boilerplate
if section.has_patterns(BF_EXACT, BF_DURATION, BF_TYPE, BF_YES, BF_FEEDING):
yield my_result(BreastfeedingStatus.BREASTFEEDING, text=section.text)
found_bf = True
if section.has_patterns(BF_NO_EXACT, BF_NOT, WHOLE_MILK):
yield my_result(BreastfeedingStatus.NO, text=section.text)
found_bf = True
if section.has_patterns(EXPRESSED_MILK_EXACT):
yield my_result(BreastfeedingStatus.EXPRESSED, text=section.text)
if section.has_patterns(BF_HISTORY):
yield my_result(BreastfeedingStatus.HISTORY, text=section.text)
if section.has_patterns(FORMULA_EXACT):
yield my_result(BreastfeedingStatus.FORMULA, text=section.text)
if section.has_patterns(FORMULA_NO):
yield my_result(BreastfeedingStatus.NO_FORMULA, text=section.text)
if section.has_patterns(PUMPING_EXACT, PUMPING_ACTIVE):
yield my_result(BreastfeedingStatus.PUMPING, text=section.text)
if section.has_patterns(BOTTLE_EXACT):
yield my_result(BreastfeedingStatus.BOTTLE, text=section.text)
if section.has_patterns(BF_STOP):
if section.has_patterns(AGO): # "stopped 2 months ago"
yield my_result(BreastfeedingStatus.STOP, text=section.text)
elif section.has_patterns(BF_STOP_BAD): # "stopped at 2 months age"
yield my_result(BreastfeedingStatus.STOPPED_BEFORE, text=section.text)
else: # "stopped"
yield my_result(BreastfeedingStatus.STOP, text=section.text)
found_bf = True
if section.has_patterns(LATCHING):
yield my_result(BreastfeedingStatus.BREASTFEEDING, text=section.text)
found_bf = True
if not found_bf and not has_boilerplate:
# boilerplate: there is at least some template language
# only non-boilerplate
if section.has_patterns(NIPPLE_SHIELD, BF_SUPPLEMENT):
yield my_result(BreastfeedingStatus.BREASTFEEDING, text=section.text)
if section.has_pattern(BF_FEEDING, ignore_negation=True):
yield my_result(BreastfeedingStatus.MAYBE, text=section.text)
cnt = section.has_patterns(BREAST_MILK, BF, PUMPING,
EXPRESSED_MILK, MILK_TRANSFER,
BREAST_PAIN,
get_count=True)
if cnt:
non_exact_count += cnt
non_exact_count_snippets.append(section.text)
if section.has_patterns(LACTATION_VISIT):
yield my_result(BreastfeedingStatus.LACTATION_VISIT, text=section.text)
if non_exact_count >= 2:
yield my_result(BreastfeedingStatus.MAYBE, text='\n'.join(non_exact_count_snippets))
def test_gNpNNNN():
doc = Document(None, text='G 1 P 1001')
status, capt, source = determine_parity(doc)
assert status == ParityStatus.P1
assert source == ParitySource.PARITY
def test_g_lt_p():
doc = Document(None, text='G 3 P 7')
status, capt, source = determine_parity(doc)
assert status == ParityStatus.SKIP
assert source == ParitySource.NONE
def test_g_eq_p_plus_1():
doc = Document(None, text='G 1 P 2')
status, capt, source = determine_parity(doc)
assert status == ParityStatus.P1
assert source == ParitySource.GRAVIDA
def test_clean_breastfeeding_document():
text = 'Breastfeeding?\nYes'
exp_text = 'Breastfeeding: Yes'
doc = Document(None, text=text)
assert doc.new_text == exp_text
def test_teaching_boilerplate_document():
text = 'Teaching/Guidance provided:\nNutrition: \nwhole milk'
exp_text = 'Teaching/Guidance provided: Nutrition: whole milk'
doc = Document(None, text=text)
assert doc.new_text == exp_text