Pymedext annotators for the EDS pipeline

Installation

Requires the installation of PyMedExt_core PyMedExt_core It can be done using requirements.txt

pip install -r requirements.txt

Installation via pip:

pip install git+git://github.com/equipe22/pymedext_eds.git@master#egg=pymedext_eds

Cloning the repository:

git clone https://github.com/equipe22/pymedext_eds.git
cd pymedext_eds
pip install .

Basic usage

All the annotators are defined in the pymedext_eds.annotators module. You will find a description of the existing annotators in the next section.

• First, import the annotators and text :

from pymedext_eds.utils import rawtext_loader

from pymedext_eds.annotators import Endlines, SentenceTokenizer, \
                                    RegexMatcher, Pipeline

from pymedext_eds.viz import display_annotations

• Load documents:

data_path = pkg_resources.resource_filename('pymedext_eds', 'data/demo')
file_list = glob(data_path + '/*.txt')
docs = [rawtext_loader(x) for x in file_list]

• Declare the pipeline:

endlines = Endlines(['raw_text'], 'endlines', 'endlines:v1')
sentences = SentenceTokenizer(['endlines'], 'sentence', 'sentenceTokenizer:v1')
regex = RegexMatcher(['endlines','syntagme'], 'regex', 'RegexMatcher:v1', 'list_regexp.json')

pipeline = Pipeline(pipeline = [endlines, sentences, regex])

• Use the pipeline to annotate:

annotated_docs = pipeline.annotate(docs)

• Explore annotations by type :

from pprint import pprint
pprint(annotated_docs[0].get_annotations('regex')[10].to_dict())

• Display annotations in text (using displacy)

display_annotations(chunk[0], ['regex'])

Existing annotators

• Endlines:
  • Used to clean the text when using text extracted from PDFs. Removes erroneous endlines introduced by pdf to text conversion.
  • input : raw_text
  • output: Annotations
• SectionSplitter:
  • Segments the text into sections
  • output: Annotations
• SentenceTokenizer:
  • Tokenize the text in sentences
  • input: cleaned text from Endlines or sections
  • output: Annotations
• Hypothesis:
  • Classification of sentences regarding the degree of certainty
  • input: sentences
  • output: Attributes
• ATCDFamille:
  • Classification of sentences regarding the subject (patient or family)
  • input: sentences
  • output: Attributes
• SyntagmeTokenizer:
  • Segmentation of sentences into syntagms
  • input: sentences
  • output: Annotations
• Negation:
  • Classification of syntagms according to the polarity
  • input: syntagm
  • output: Attributes
• RegexMatcher:
  • Extracts informations using predefined regexs
  • input: sentence or syntagm
  • output: Annotations
• QuickUMLSAnnotator:
  • Extracts medical concepts from UMLS using QuickUMLS
  • output: Annotations
• MedicationAnnotator:
  • Extracts medications informations using a deep learning pipeline
  • output: Annotations

QuickUMLS installation (copied from Georgetown-IR-Lab/QuickUMLS)

Installation

1. Obtain a UMLS installation This tool requires you to have a valid UMLS installation on disk. To install UMLS, you must first obtain a license from the National Library of Medicine; then you should download all UMLS files from this page; finally, you can install UMLS using the MetamorphoSys tool as explained in this guide. The installation can be removed once the system has been initialized.
2. Install QuickUMLS: You can do so by either running pip install quickumls or python setup.py install. On macOS, using anaconda is strongly recommended†.
3. Create a QuickUMLS installation Initialize the system by running python -m quickumls.install <umls_installation_path> <destination_path>, where <umls_installation_path> is where the installation files are (in particular, we need MRCONSO.RRF and MRSTY.RRF) and <destination_path> is the directory where the QuickUmls data files should be installed. This process will take between 5 and 30 minutes depending how fast the CPU and the drive where UMLS and QuickUMLS files are stored are (on a system with a Intel i7 6700K CPU and a 7200 RPM hard drive, initialization takes 8.5 minutes).

python -m quickumls.install supports the following optional arguments:

• -L / --lowercase: if used, all concept terms are folded to lowercase before being processed. This option typically increases recall, but it might reduce precision;
• -U / --normalize-unicode: if used, expressions with non-ASCII characters are converted to the closest combination of ASCII characters.
• -E / --language: Specify the language to consider for UMLS concepts; by default, English is used. For a complete list of languages, please see this table provided by NLM.
• -d / --database-backend: Specify which database backend to use for QuickUMLS. The two options are leveldb and unqlite. The latter supports multi-process reading and has better unicode compatibility, and it used as default for all new 1.4 installations; the former is still used as default when instantiating a QuickUMLS client. More info about differences between the two databases and migration info are available here.

†: If the installation fails on macOS when using Anaconda, install leveldb first by running conda install -c conda-forge python-leveldb.

Run a simple server

Define the server and the pipeline:

import flask

from flask import Flask, render_template, request

from pymedext_eds.annotators import Endlines, SentenceTokenizer, Hypothesis, \
                                    ATCDFamille, SyntagmeTokenizer, Negation, RegexMatcher, \
                                    Pipeline

endlines = Endlines(['raw_text'], 'endlines', 'endlines:v1')
sentences = SentenceTokenizer(['endlines'], 'sentence', 'sentenceTokenizer:v1')
hypothesis = Hypothesis(['sentence'], 'hypothesis', 'hypothesis:v1')
family = ATCDFamille(['sentence'], 'context', 'ATCDfamily:v1')
syntagmes = SyntagmeTokenizer(['sentence'], 'syntagme', 'SyntagmeTokenizer:v1')
negation = Negation(['syntagme'], 'negation', 'Negation:v1')
regex = RegexMatcher(['endlines','syntagme'], 'regex', 'RegexMatcher:v1', 'list_regexp.json')

pipeline = Pipeline(pipeline = [endlines, sentences, hypothesis, family, syntagmes, negation, regex])

app=Flask(__name__)

@app.route('/annotate',methods = ['POST'])
def result():
    if request.method == 'POST':

        return pipeline.__call__(request)

if __name__ == '__main__':
    app.run(port = 6666, debug=True)

Save this code in demo_flask_server.py and run it using:

python demo_flask_server.py

Query the server:

import requests
from pymedextcore.document import Document

data_path = pkg_resources.resource_filename('pymedext_eds', 'data/demo')
file_list = glob(data_path + '/*.txt')
docs = [rawtext_loader(x) for x in file_list]

json_doc = [doc.to_dict() for doc in docs]
res =  requests.post(f"http://127.0.0.1:6666/annotate", json = json_doc)
if res.status_code == 200:
    res = res.json()['result']
    docs = [Document.from_dict(doc) for doc in res ]

Run a docker server

define the git credentials

first create a file .git-credentials and replace user and pass by your github credentials such has

https://user:pass@github.com

WARNING :never add it on the git !!!

build the images

docker build -f eds_apps/Dockerfile_backend -t pymedext-eds:v1 .


#if proxy add
docker build -f eds_apps/Dockerfile_backend -t pymedext-eds:v1 \
--buildargs http_proxy="proxy" \
--buildargs https_proxy="proxy" .

start the backend server

docker run --rm  -d -p 6666:6666 pymedext-eds:v1 python3 demo_flask.py