Description
Deidentification NER (Large) is a Named Entity Recognition model that annotates text to find protected health information that may need to be deidentified. The entities it annotates are Age, Contact, Date, Id, Location, Name, and Profession. This model is trained with the 'embeddings_clinical' word embeddings model, so be sure to use the same embeddings in the pipeline.
We sticked to official annotation guideline (AG) for 2014 i2b2 Deid challenge while annotating new datasets for this model. All the details regarding the nuances and explanations for AG can be found here https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4978170/
Predicted Entities
AGE, CONTACT, DATE, ID, LOCATION, NAME, PROFESSION, HEALTHPLAN, URL.
Live Demo Open in Colab Copy S3 URI
How to use
document_assembler = DocumentAssembler() \
.setInputCol("text") \
.setOutputCol("document")
sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl_healthcare","en","clinical/models")\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer() \
.setInputCols("sentence") \
.setOutputCol("token")
word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")
ner = MedicalNerModel.pretrained("ner_deid_large","en","clinical/models") \
.setInputCols("sentence","token","embeddings") \
.setOutputCol("ner")
ner_converter = NerConverter() \
.setInputCols(["sentence", "token", "ner"]) \
.setOutputCol("entities")
nlp_pipeline = Pipeline(stages=[
document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
ner,
ner_converter])
light_pipeline = LightPipeline(nlp_pipeline.fit(spark.createDataFrame([['']]).toDF("text")))
input_text = ["""HISTORY OF PRESENT ILLNESS: Mr. Smith is a 60-year-old white male veteran with multiple comorbidities, who has a history of bladder cancer diagnosed approximately two years ago by the VA Hospital. He underwent a resection there. He was to be admitted to the Day Hospital for cystectomy. He was seen in Urology Clinic and Radiology Clinic on 02/04/2003. HOSPITAL COURSE: Mr. Smith presented to the Day Hospital in anticipation for Urology surgery. On evaluation, EKG, echocardiogram was abnormal, a Cardiology consult was obtained. A cardiac adenosine stress MRI was then proceeded, same was positive for inducible ischemia, mild-to-moderate inferolateral subendocardial infarction with peri-infarct ischemia. In addition, inducible ischemia seen in the inferior lateral septum. Mr. Smith underwent a left heart catheterization, which revealed two vessel coronary artery disease. The RCA, proximal was 95% stenosed and the distal 80% stenosed. The mid LAD was 85% stenosed and the distal LAD was 85% stenosed. There was four Multi-Link Vision bare metal stents placed to decrease all four lesions to 0%. Following intervention, Mr. Smith was admitted to 7 Ardmore Tower under Cardiology Service under the direction of Dr. Hart. Mr. Smith had a noncomplicated post-intervention hospital course. He was stable for discharge home on 02/07/2003 with instructions to take Plavix daily for one month and Urology is aware of the same."""]
result = pipeline_model.transform(spark.createDataFrame([input_text], ["text"]))
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl_healthcare","en","clinical/models")
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val ner = MedicalNerModel.pretrained("ner_deid_large","en","clinical/models")
.setInputCols(Array("sentence","token","embeddings"))
.setOutputCol("ner")
ner_converter = new NerConverter()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("entities")
val pipeline = new Pipeline().setStages(Array(
document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
ner,
ner_converter))
val data = Seq("HISTORY OF PRESENT ILLNESS: Mr. Smith is a 60-year-old white male veteran with multiple comorbidities, who has a history of bladder cancer diagnosed approximately two years ago by the VA Hospital. He underwent a resection there. He was to be admitted to the Day Hospital for cystectomy. He was seen in Urology Clinic and Radiology Clinic on 02/04/2003. HOSPITAL COURSE: Mr. Smith presented to the Day Hospital in anticipation for Urology surgery. On evaluation, EKG, echocardiogram was abnormal, a Cardiology consult was obtained. A cardiac adenosine stress MRI was then proceeded, same was positive for inducible ischemia, mild-to-moderate inferolateral subendocardial infarction with peri-infarct ischemia. In addition, inducible ischemia seen in the inferior lateral septum. Mr. Smith underwent a left heart catheterization, which revealed two vessel coronary artery disease. The RCA, proximal was 95% stenosed and the distal 80% stenosed. The mid LAD was 85% stenosed and the distal LAD was 85% stenosed. There was four Multi-Link Vision bare metal stents placed to decrease all four lesions to 0%. Following intervention, Mr. Smith was admitted to 7 Ardmore Tower under Cardiology Service under the direction of Dr. Hart. Mr. Smith had a noncomplicated post-intervention hospital course. He was stable for discharge home on 02/07/2003 with instructions to take Plavix daily for one month and Urology is aware of the same.").toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("en.med_ner.deid.large").predict("""HISTORY OF PRESENT ILLNESS: Mr. Smith is a 60-year-old white male veteran with multiple comorbidities, who has a history of bladder cancer diagnosed approximately two years ago by the VA Hospital. He underwent a resection there. He was to be admitted to the Day Hospital for cystectomy. He was seen in Urology Clinic and Radiology Clinic on 02/04/2003. HOSPITAL COURSE: Mr. Smith presented to the Day Hospital in anticipation for Urology surgery. On evaluation, EKG, echocardiogram was abnormal, a Cardiology consult was obtained. A cardiac adenosine stress MRI was then proceeded, same was positive for inducible ischemia, mild-to-moderate inferolateral subendocardial infarction with peri-infarct ischemia. In addition, inducible ischemia seen in the inferior lateral septum. Mr. Smith underwent a left heart catheterization, which revealed two vessel coronary artery disease. The RCA, proximal was 95% stenosed and the distal 80% stenosed. The mid LAD was 85% stenosed and the distal LAD was 85% stenosed. There was four Multi-Link Vision bare metal stents placed to decrease all four lesions to 0%. Following intervention, Mr. Smith was admitted to 7 Ardmore Tower under Cardiology Service under the direction of Dr. Hart. Mr. Smith had a noncomplicated post-intervention hospital course. He was stable for discharge home on 02/07/2003 with instructions to take Plavix daily for one month and Urology is aware of the same.""")
Results
+---------------+---------+
|chunk |ner_label|
+---------------+---------+
|Smith |NAME |
|VA Hospital |LOCATION |
|Day Hospital |LOCATION |
|02/04/2003 |DATE |
|Smith |NAME |
|Day Hospital |LOCATION |
|Smith |NAME |
|Smith |NAME |
|7 Ardmore Tower|LOCATION |
|Hart |NAME |
|Smith |NAME |
|02/07/2003 |DATE |
+---------------+---------+
Model Information
| Model Name: | ner_deid_large |
| Compatibility: | Healthcare NLP 3.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | en |
Data Source
Trained on plain n2c2 2014: De-identification and Heart Disease Risk Factors Challenge datasets with embeddings_clinical
https://portal.dbmi.hms.harvard.edu/projects/n2c2-2014/
Benchmarking
label tp fp fn prec rec f1
I-NAME 1096 47 80 0.95888 0.931973 0.945235
I-CONTACT 93 0 4 1 0.958763 0.978947
I-AGE 3 1 6 0.75 0.333333 0.461538
B-DATE 2078 42 52 0.980189 0.975587 0.977882
I-DATE 474 39 25 0.923977 0.9499 0.936759
I-LOCATION 755 68 76 0.917375 0.908544 0.912938
I-PROFESSION 78 8 9 0.906977 0.896552 0.901734
B-NAME 1182 101 36 0.921278 0.970443 0.945222
B-AGE 259 10 11 0.962825 0.959259 0.961039
B-ID 146 8 11 0.948052 0.929936 0.938907
B-PROFESSION 76 9 21 0.894118 0.783505 0.835165
B-LOCATION 556 87 71 0.864697 0.886762 0.875591
I-ID 64 8 3 0.888889 0.955224 0.920863
B-CONTACT 40 7 5 0.851064 0.888889 0.869565
Macro-average 6900 435 410 0.912023 0.880619 0.896046
Micro-average 6900 435 410 0.940695 0.943912 0.942301