Description
Assign assertion status to clinical entities extracted by NER based on their context in the text.
Predicted Entities
absent, present, conditional, associated_with_someone_else, hypothetical, possible.
Live Demo Open in Colab Copy S3 URI
How to use
documentAssembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentenceDetector = SentenceDetector()\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
word_embeddings = BertEmbeddings.pretrained("biobert_pubmed_base_cased")\
.setInputCols(["sentence",'token'])\
.setOutputCol("embeddings")
clinical_ner = MedicalNerModel.pretrained("ner_clinical_biobert", "en", "clinical/models") \
.setInputCols(["sentence", "token", "embeddings"]) \
.setOutputCol("ner")
ner_converter = NerConverter() \
.setInputCols(["sentence", "token", "ner"]) \
.setOutputCol("ner_chunk")
clinical_assertion = AssertionDLModel.pretrained("assertion_dl_biobert", "en", "clinical/models") \
.setInputCols(["sentence", "ner_chunk", "embeddings"]) \
.setOutputCol("assertion")
nlpPipeline = Pipeline(stages=[
documentAssembler,
sentenceDetector,
tokenizer,
word_embeddings,
clinical_ner,
ner_converter,
clinical_assertion
])
data = spark.createDataFrame([["""Patient has a headache for the last 2 weeks, needs to get a head CT, and appears anxious when she walks fast. No alopecia noted. She denies pain"""]]).toDF("text")
result = nlpPipeline.fit(data).transform(data)
val documentAssembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentenceDetector = new SentenceDetector()
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val word_embeddings = BertEmbeddings.pretrained("biobert_pubmed_base_cased")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
clinical_ner = MedicalNerModel.pretrained("ner_clinical_biobert", "en", "clinical/models")
.setInputCols(Array("sentence", "token", "embeddings"))
.setOutputCol("ner")
val ner_converter = new NerConverter()
.setInputCols(Array("sentence","token","ner"))
.setOutputCol("ner_chunk")
val clinical_assertion = AssertionDLModel.pretrained("assertion_dl_biobert","en", "clinical/models")
.setInputCols(Array("sentence", "ner_chunk", "embeddings"))
.setOutputCol("assertion")
val pipeline = new Pipeline().setStages(Array(documentAssembler,
sentenceDetector,
tokenizer,
word_embeddings,
clinical_ner,
ner_converter,
clinical_assertion))
val data = Seq("""Patient has a headache for the last 2 weeks, needs to get a head CT, and appears anxious when she walks fast. No alopecia noted. She denies pain""").toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("en.assert.biobert").predict("""Patient has a headache for the last 2 weeks, needs to get a head CT, and appears anxious when she walks fast. No alopecia noted. She denies pain.""")
Results
+----------+---------+-----------+
|chunk |ner_label|assertion |
+----------+---------+-----------+
|a headache|PROBLEM |present |
|a head CT |TEST |present |
|anxious |PROBLEM |conditional|
|alopecia |PROBLEM |absent |
|pain |PROBLEM |absent |
+----------+---------+-----------+
Model Information
| Model Name: | assertion_dl_biobert |
| Compatibility: | Spark NLP 2.7.2+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [document, chunk, embeddings] |
| Output Labels: | [assertion] |
| Language: | en |
Data Source
Trained with augmented version of i2b2 dataset.
Benchmarking
label tp fp fn prec rec f1
absent 769 51 57 0.937805 0.930993 0.934386
present 2575 161 102 0.941155 0.961898 0.951413
conditional 20 14 23 0.588235 0.465116 0.519481
associated_with_someone_else 51 9 15 0.85 0.772727 0.809524
hypothetical 129 13 15 0.908451 0.895833 0.902098
possible 114 44 80 0.721519 0.587629 0.647727
Macro-average 3658 292 292 0.824527 0.769033 0.795814
Micro-average 3658 292 292 0.926076 0.926076 0.926076