Detect PHI for Deidentification purposes (Spanish)

Description

Named Entity Recognition annotators allow for a generic model to be trained by using a Deep Learning architecture (Char CNNs - BiLSTM - CRF - word embeddings) inspired on a former state of the art model for NER: Chiu & Nicols, Named Entity Recognition with Bidirectional LSTM,CNN.

Deidentification NER (Spanish) is a Named Entity Recognition model that annotates text to find protected health information that may need to be de-identified. It detects 13 entities. This NER model is trained with a combination of custom datasets, Spanish 2002 conLL, MeddoProf dataset and several data augmentation mechanisms.

Predicted Entities

PATIENT, HOSPITAL, DATE, ORGANIZATION, E-MAIL, USERNAME, LOCATION, ZIP, MEDICALRECORD, PROFESSION, PHONE, DOCTOR, AGE

Live Demo Open in Colab Download Copy S3 URI

How to use

documentAssembler = nlp.DocumentAssembler()\
	.setInputCol("text")\
	.setOutputCol("document")

sentenceDetector = nlp.SentenceDetectorDLModel.pretrained("sentence_detector_dl","xx")\
	.setInputCols(["document"])\
	.setOutputCol("sentence")

tokenizer = nlp.Tokenizer()\
	.setInputCols(["sentence"])\
	.setOutputCol("token")

embeddings = nlp.WordEmbeddingsModel.pretrained("embeddings_sciwiki_300d","es","clinical/models")\
	.setInputCols(["sentence","token"])\
	.setOutputCol("word_embeddings")

clinical_ner = medical.NerModel.pretrained("ner_deid_subentity", "es", "clinical/models")\
	.setInputCols(["sentence","token","word_embeddings"])\
	.setOutputCol("ner")

nlpPipeline = nlp.Pipeline(stages=[
	documentAssembler,
	sentenceDetector,
	tokenizer,
	embeddings,
	clinical_ner])

text = ['''
Antonio Pérez Juan, nacido en Cadiz, España. Aún no estaba vacunado, se infectó con Covid-19 el dia 14/03/2020 y tuvo que ir al Hospital. Fue tratado con anticuerpos monoclonales en la Clinica San Carlos.
''']

data = spark.createDataFrame([text]).toDF("text")

results = nlpPipeline.fit(data).transform(data)

val documentAssembler = new DocumentAssembler()
	.setInputCol("text")
	.setOutputCol("document")

val sentenceDetector = SentenceDetectorDLModel.pretrained("sentence_detector_dl_healthcare","xx")
	.setInputCols(Array("document"))
	.setOutputCol("sentence")

val tokenizer = new Tokenizer()
	.setInputCols(Array("sentence"))
	.setOutputCol("token")

val embeddings = WordEmbeddingsModel.pretrained("embeddings_sciwiki_300d","es","clinical/models")
	.setInputCols(Array("sentence", "token"))
	.setOutputCol("embeddings")

val clinical_ner = MedicalNerModel.pretrained("ner_deid_subentity", "es", "clinical/models")
	.setInputCols(Array("sentence","token","embeddings"))
	.setOutputCol("ner")

val pipeline = new Pipeline().setStages(Array(
	documentAssembler, 
	sentenceDetector, 
	tokenizer, 
	embeddings,
	clinical_ner))

val text = """Antonio Pérez Juan, nacido en Cadiz, España. Aún no estaba vacunado, se infectó con Covid-19 el dia 14/03/2020 y tuvo que ir al Hospital. Fue tratado con anticuerpos monoclonales en la Clinica San Carlos."""

val data = Seq(text).toDS.toDF("text")

val results = pipeline.fit(data).transform(data)

import nlu
nlu.load("es.med_ner.deid.subentity").predict("""
Antonio Pérez Juan, nacido en Cadiz, España. Aún no estaba vacunado, se infectó con Covid-19 el dia 14/03/2020 y tuvo que ir al Hospital. Fue tratado con anticuerpos monoclonales en la Clinica San Carlos.
""")

Results

+------------+----------+
|       token| ner_label|
+------------+----------+
|     Antonio| B-PATIENT|
|       Pérez| I-PATIENT|
|        Juan| I-PATIENT|
|           ,|         O|
|      nacido|         O|
|          en|         O|
|       Cadiz|B-LOCATION|
|           ,|         O|
|      España|B-LOCATION|
|           .|         O|
|         Aún|         O|
|          no|         O|
|      estaba|         O|
|    vacunado|         O|
|           ,|         O|
|          se|         O|
|     infectó|         O|
|         con|         O|
|    Covid-19|         O|
|          el|         O|
|         dia|         O|
|          14|    B-DATE|
|          de|    I-DATE|
|       Marzo|    I-DATE|
|           y|         O|
|        tuvo|         O|
|         que|         O|
|          ir|         O|
|          al|         O|
|    Hospital|         O|
|         Fue|         O|
|     tratado|         O|
|         con|         O|
| anticuerpos|         O|
|monoclonales|         O|
|          en|         O|
|          la|         O|
|     Clinica|B-HOSPITAL|
|         San|I-HOSPITAL|
|      Carlos|I-HOSPITAL|
|           .|         O|
+------------+----------+

Model Information

Model Name:	ner_deid_subentity
Compatibility:	Healthcare NLP 3.3.4+
License:	Licensed
Edition:	Official
Input Labels:	[sentence, token, word_embeddings]
Output Labels:	[ner]
Language:	es
Size:	15.0 MB
Dependencies:	embeddings_sciwiki_300d

Data Source

Internal JSL annotated corpus
Spanish conLL
MeddoProf

Benchmarking

label      tp     fp     fn   total  precision  recall      f1
PATIENT  2088.0  201.0  178.0  2266.0     0.9122  0.9214  0.9168
HOSPITAL   302.0   43.0   85.0   387.0     0.8754  0.7804  0.8251
DATE  1837.0   33.0   20.0  1857.0     0.9824  0.9892  0.9858
ORGANIZATION  2498.0  477.0  649.0  3147.0     0.8397  0.7938  0.8161
MAIL    58.0    0.0    0.0    58.0        1.0     1.0     1.0
USERNAME    90.0    0.0   15.0   105.0        1.0  0.8571  0.9231
LOCATION  1866.0  391.0  354.0  2220.0     0.8268  0.8405  0.8336
ZIP    20.0    1.0    2.0    22.0     0.9524  0.9091  0.9302
MEDICALRECORD   111.0    5.0   20.0   131.0     0.9569  0.8473  0.8988
PROFESSION   270.0   96.0  134.0   404.0     0.7377  0.6683  0.7013
PHONE   108.0   11.0    8.0   116.0     0.9076   0.931  0.9191
DOCTOR   659.0   40.0   40.0   699.0     0.9428  0.9428  0.9428
AGE   302.0   53.0   61.0   363.0     0.8507   0.832  0.8412
macro     -      -      -       -         -      -      0.8872247
micro     -      -      -       -         -      -      0.8741892

PREVIOUSDetect PHI for Deidentification purposes (Spanish, reduced entities)

NEXTNER Model Finder