Skip to content

Score

The eds.score pipeline allows easy extraction of typical scores (Charlson, SOFA...) that can be found in clinical documents. The pipeline works by

  • Extracting the score's name via the provided regular expressions
  • Extracting the score's raw value via another set of RegEx
  • Normalising the score's value via a normalising function

Charlson Comorbidity Index

Implementing the eds.score pipeline, the charlson pipeline will extract the Charlson Comorbidity Index:

import spacy

nlp = spacy.blank("fr")
nlp.add_pipe("eds.sentences")
nlp.add_pipe("eds.normalizer")
nlp.add_pipe("eds.charlson")

text = "Charlson à l'admission: 7.\n" "Charlson: \n" "OMS: \n"

doc = nlp(text)
doc.ents
# Out: (Charlson à l'admission: 7,)

We can see that only one occurrence was extracted. The second mention of Charlson in the text doesn't contain any numerical value, so it isn't extracted.

Each extraction exposes 2 extensions:

ent = doc.ents[0]

ent._.score_name
# Out: 'eds.charlson'

ent._.score_value
# Out: 7

SOFA score

The SOFA pipe allows to extract SOFA scores.

import spacy

nlp = spacy.blank("fr")
nlp.add_pipe("eds.sentences")
nlp.add_pipe("eds.normalizer")
nlp.add_pipe("eds.SOFA")

text = "SOFA (à 24H) : 12.\n" "OMS: \n"

doc = nlp(text)
doc.ents
# Out: (SOFA (à 24H) : 12,)

Each extraction exposes 3 extensions:

ent = doc.ents[0]

ent._.score_name
# Out: 'eds.SOFA'

ent._.score_value
# Out: 12

ent._.score_method
# Out: '24H'

Score method can here be "24H", "Maximum", "A l'admission" or "Non précisée"

TNM score

The eds.TNM pipe allows to extract TNM scores.

import spacy

nlp = spacy.blank("fr")
nlp.add_pipe("eds.sentences")
nlp.add_pipe("eds.TNM")

text = "TNM: pTx N1 M1"

doc = nlp(text)
doc.ents
# Out: (pTx N1 M1,)

ent = doc.ents[0]
ent._.value.dict()
# {'modifier': 'p',
#  'tumour': None,
#  'tumour_specification': 'x',
#  'node': '1',
#  'node_specification': None,
#  'metastasis': '1',
#  'resection_completeness': None,
#  'version': None,
#  'version_year': None}

The TNM score is based on the developement of S. Priou, B. Rance and E. Kempf 1.

Implementing your own score

Using the eds.score pipeline, you only have to change its configuration in order to implement a simple score extraction algorithm. As an example, let us see the configuration used for the eds.charlson pipe The configuration consists of 4 items:

  • score_name: The name of the score
  • regex: A list of regular expression to detect the score's mention
  • value_extract: A regular expression to extract the score's value in the context of the score's mention
  • score_normalization: A function name used to normalise the score's raw value

Note

spaCy doesn't allow to pass functions in the configuration of a pipeline. To circumvent this issue, functions need to be registered, which simply consists in decorating those functions

The registration is done as follows:

@spacy.registry.misc("score_normalization.charlson")
def my_normalization_score(raw_score: str):
    # Implement some filtering here
    # Return None if you want the score to be discarded
    return normalized_score

The values used for the eds.charlson pipe are the following:

@spacy.registry.misc("score_normalization.charlson")
def score_normalization(extracted_score):
    """
    Charlson score normalization.
    If available, returns the integer value of the Charlson score.
    """
    score_range = list(range(0, 30))
    if (extracted_score is not None) and (int(extracted_score) in score_range):
        return int(extracted_score)


charlson_config = dict(
    score_name="charlson",
    regex=[r"charlson"],
    value_extract=r"charlson.*[\n\W]*(\d+)",
    score_normalization="score_normalization.charlson",
)

  1. Emmanuelle Kempf, Sonia Priou, Guillaume Lamé, Christel Daniel, Ali Bellamine, Daniele Sommacale, yazid Belkacemi, Romain Bey, Gilles Galula, Namik Taright, Xavier Tannier, Bastien Rance, Rémi Flicoteaux, François Hemery, Etienne Audureau, Gilles Chatellier, and Christophe Tournigand. Impact of two waves of Sars-Cov2 outbreak on the number, clinical presentation, care trajectories and survival of patients newly referred for a colorectal cancer: A French multicentric cohort study from a large group of University hospitals. International Journal of Cancer, 150(10):1609–1618, 2022. URL: https://hal.archives-ouvertes.fr/hal-03519085, doi:10.1002/ijc.33928. ↩