EDS-Medic

EDS-Medic is a project used at APHP to extract drug prescriptions from clinical reports. It is built on top of edsnlp and spaCy.

Installation

pip install eds-medic

# To use with a GPU, you will need to specify your version of CUDA
pip install 'eds-medic[cuda110]'

Loading the model

import eds_medic

nlp = eds_medic.load()
# add components to the model and/or run it on documents

Training a model

We created a single workflow that:

• Converts the datasets from JSONL or BRAT files to spaCy format
• Trains the pipeline
• Evaluates the pipeline using the test set
• Packages the resulting model to make it pip-installable

To use it, you will need to supply:

• A labelled dataset
• A HuggingFace transformers model, or use camembert-base

In any case, you will need to modify the configuration to reflect these changes.

Entities and attributes

The default model is trained to recognize the following entities:

Entity	Description
drug	the drug name
class	a active substance class
dose	the dose of the drug
frequency	a frequency for the drug intake
criteria	a criteria for the drug intake
duration	how long the prescription is valid
route	the route of the drug intake

And extract the following attributes:

Attribute	Description	Entities
ent._.event_type	the event type	drug, class, date

Commands

Command	Description
convert	Convert data to spaCy's binary format
train	Train the NER model
evaluate	Evaluate the model and export metrics
package	Package the trained model as a pip package
visualize-model	Visualize the model's output interactively using Streamlit

Run the command with

spacy project run [command] [options]

Documentation

Visit the documentation for more information!

Acknowledgement

We would like to thank Assistance Publique – Hôpitaux de Paris and AP-HP Foundation for funding this project.