Description
This is a BERT-based model for classification of clinical documents sections. This model performs better when the section header is present in the text, e.g., when splitting the document with ChunkSentenceSplitter annotator with parameter setInsertChunk=True.
Predicted Entities
Complications and Risk Factors, Consultation and Referral, Diagnostic and Laboratory Data, Discharge Information, Habits, History, Patient Information, Procedures, Impression, Other
How to use
document_assembler = nlp.DocumentAssembler() \
.setInputCol("text") \
.setOutputCol("document")
tokenizer = nlp.Tokenizer() \
.setInputCols(["document"]) \
.setOutputCol("token")
sequenceClassifier = medical.BertForSequenceClassification.pretrained("bert_sequence_classifier_clinical_sections", "en", "clinical/models")\
.setInputCols(["document", "token"])\
.setOutputCol("prediction")\
.setCaseSensitive(False)
pipeline = nlp.Pipeline(stages=[
document_assembler,
tokenizer,
sequenceClassifier
])
example_df = spark.createDataFrame(
[["""Discharge Instructions:
It was a pleasure taking care of you! You came to us with
stomach pain and worsening distension. While you were here we
did a paracentesis to remove 1.5L of fluid from your belly. We
also placed you on you 40 mg of Lasix and 50 mg of Aldactone to
help you urinate the excess fluid still in your belly. As we
discussed, everyone has a different dose of lasix required to
make them urinate and it's likely that you weren't taking a high
enough dose. Please take these medications daily to keep excess
fluid off and eat a low salt diet. You will follow up with Dr.
___ in liver clinic and from there have your colonoscopy
and EGD scheduled. """]]).toDF("text")
result = spark_model.transform(example_df)
result.select("prediction.result").show(truncate=False)
val documentAssembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentenceDetector = SentenceDetectorDLModel.pretrained()
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val seq = BertForSequenceClassification.pretrained("bert_sequence_classifier_clinical_sections", "en", "clinical/models")
.setInputCols(Array("token", "sentence"))
.setOutputCol("label")
.setCaseSensitive(false)
val pipeline = new Pipeline().setStages(Array(
documentAssembler,
sentenceDetector,
tokenizer,
seq))
val test_sentences = """Discharge Instructions:
It was a pleasure taking care of you! You came to us with
stomach pain and worsening distension. While you were here we
did a paracentesis to remove 1.5L of fluid from your belly. We
also placed you on you 40 mg of Lasix and 50 mg of Aldactone to
help you urinate the excess fluid still in your belly. As we
discussed, everyone has a different dose of lasix required to
make them urinate and it's likely that you weren't taking a high
enough dose. Please take these medications daily to keep excess
fluid off and eat a low salt diet. You will follow up with Dr.
___ in liver clinic and from there have your colonoscopy
and EGD scheduled. """"
val example = Seq(test_sentences).toDF("text")
val result = pipeline.fit(example).transform(example)
import nlu
nlu.load("en.classify.bert_sequence.clinical_sections").predict("""Discharge Instructions:
It was a pleasure taking care of you! You came to us with
stomach pain and worsening distension. While you were here we
did a paracentesis to remove 1.5L of fluid from your belly. We
also placed you on you 40 mg of Lasix and 50 mg of Aldactone to
help you urinate the excess fluid still in your belly. As we
discussed, everyone has a different dose of lasix required to
make them urinate and it's likely that you weren't taking a high
enough dose. Please take these medications daily to keep excess
fluid off and eat a low salt diet. You will follow up with Dr.
___ in liver clinic and from there have your colonoscopy
and EGD scheduled. """)
Results
+-----------------------+
|result |
+-----------------------+
|[Discharge Information]|
+-----------------------+
Model Information
| Model Name: | bert_sequence_classifier_clinical_sections |
| Compatibility: | Healthcare NLP 5.1.4+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [document, token] |
| Output Labels: | [class] |
| Language: | en |
| Size: | 406.6 MB |
| Case sensitive: | false |
| Max sentence length: | 512 |
References
In-house annotation of clinical documents.
Sample text from the training dataset
Discharge Instructions: It was a pleasure taking care of you! You came to us with stomach pain and worsening distension. While you were here we did a paracentesis to remove 1.5L of fluid from your belly. We also placed you on you 40 mg of Lasix and 50 mg of Aldactone to help you urinate the excess fluid still in your belly. As we discussed, everyone has a different dose of lasix required to make them urinate and it’s likely that you weren’t taking a high enough dose. Please take these medications daily to keep excess fluid off and eat a low salt diet. You will follow up with Dr. ___ in liver clinic and from there have your colonoscopy and EGD scheduled.
Benchmarking
label precision recall f1-score support
Consultation_and_Referral 0.981203 0.996183 0.988636 262
Other 1.000000 1.000000 1.000000 29
Habits 0.983051 1.000000 0.991453 58
Complications_and_Risk_Factors 1.000000 1.000000 1.000000 385
Diagnostic_and_Laboratory_Data 0.987835 0.983051 0.985437 413
Discharge_Information 0.992386 0.982412 0.987374 398
History 1.000000 0.990099 0.995025 404
Impression 0.997706 0.997706 0.997706 436
Patient_Information 0.994764 0.994764 0.994764 382
Procedures 0.984456 0.997375 0.990874 381
accuracy - - 0.992694 3148
macro-avg 0.992140 0.994159 0.993127 3148
weighted-avg 0.992730 0.992694 0.992694 3148