Skip to content

Drugs

The eds.drugs pipeline component detects mentions of French drugs (brand names and active ingredients) and adds them to doc.ents. Each drug is mapped to an ATC code through the Romedi terminology 1. The ATC classifies drugs into groups.

Usage

In this example, we are looking for an oral antidiabetic medication (ATC code: A10B).

from edsnlp.pipelines.core.terminology import TerminologyTermMatcher
import spacy

nlp = spacy.blank("fr")
nlp.add_pipe("eds.normalizer")
nlp.add_pipe("eds.drugs", config=dict(term_matcher=TerminologyTermMatcher.exact))

text = "Traitement habituel: Kardégic, cardensiel (bisoprolol), glucophage, lasilix"

doc = nlp(text)

drugs_detected = [(x.text, x.kb_id_) for x in doc.ents]

drugs_detected
# Out: [('Kardégic', 'B01AC06'), ('cardensiel', 'C07AB07'), ('bisoprolol', 'C07AB07'), ('glucophage', 'A10BA02'), ('lasilix', 'C03CA01')]

oral_antidiabetics_detected = list(
    filter(lambda x: (x[1].startswith("A10B")), drugs_detected)
)
oral_antidiabetics_detected
# Out: [('glucophage', 'A10BA02')]

Glucophage is the brand name of a medication that contains metformine, the first-line medication for the treatment of type 2 diabetes.

Configuration

The pipeline can be configured using the following parameters :

Parameter Description Default
term_matcher Which algorithm should we use : exact or simstring "LOWER"
term_matcher_config Config of the algorithm (SimstringMatcher's for simstring) "LOWER"
attr spaCy attribute to match on (eg NORM, TEXT, LOWER) "NORM"
ignore_excluded Whether to ignore excluded tokens for matching False

Authors and citation

The eds.drugs pipeline was developed by the IAM team and CHU de Bordeaux's Data Science team.

  1. Sébastien Cossin, Luc Lebrun, Grégory Lobre, Romain Loustau, Vianney Jouhet, Romain Griffier, Fleur Mougin, Gayo Diallo, and Frantz Thiessard. Romedi: An Open Data Source About French Drugs on the Semantic Web. Studies in Health Technology and Informatics, 264:79–82, August 2019. URL: https://hal.archives-ouvertes.fr/hal-02987843, doi:10.3233/SHTI190187. ↩