Sentence Entity Resolver for UMLS CUI Codes

Description

This model maps clinical entities and concepts to 4 major categories of UMLS CUI codes using sbiobert_base_cased_mli Sentence Bert Embeddings. It has faster load time, with a speedup of about 6X when compared to previous versions. Also the load process now is more memory friendly meaning that the maximum memory required during load time is smaller, reducing the chances of OOM exceptions, and thus relaxing hardware requirements.

This model returns CUI (concept unique identifier) codes for 200K concepts from clinical findings.https://www.nlm.nih.gov/research/umls/index.html

Predicted Entities

umls cuı codes

Live Demo Open in Colab Copy S3 URI

How to use

...
document_assembler = DocumentAssembler()\
      .setInputCol('text')\
      .setOutputCol('document')

sentence_detector = SentenceDetector()\
      .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_model = MedicalNerModel.pretrained("ner_clinical", "en", "clinical/models")\
    .setInputCols(["sentence", "token", "embeddings"])\
    .setOutputCol("clinical_ner")

ner_model_converter = NerConverterInternal()\
    .setInputCols(["sentence", "token", "clinical_ner"])\
    .setOutputCol("ner_chunk")

chunk2doc = Chunk2Doc().setInputCols("ner_chunk").setOutputCol("ner_chunk_doc")

sbert_embedder = BertSentenceEmbeddings\
     .pretrained("sbiobert_base_cased_mli",'en','clinical/models')\
     .setInputCols(["ner_chunk_doc"])\
     .setOutputCol("sbert_embeddings")

resolver = SentenceEntityResolverModel.pretrained("sbiobertresolve_umls_findings","en", "clinical/models") \
     .setInputCols(["ner_chunk_doc", "sbert_embeddings"]) \
     .setOutputCol("resolution")\
     .setDistanceFunction("EUCLIDEAN")

pipeline = Pipeline(stages = [document_assembler, sentence_detector, tokenizer, word_embeddings, ner_model, ner_model_converter, chunk2doc, sbert_embedder, resolver])

data = spark.createDataFrame([["""A 28-year-old female with a history of gestational diabetes mellitus diagnosed eight years prior to presentation and subsequent type two diabetes mellitus (T2DM), one prior episode of HTG-induced pancreatitis three years prior to presentation, associated with an acute hepatitis, and obesity with a body mass index (BMI) of 33.5 kg/m2, presented with a one-week history of polyuria, polydipsia, poor appetite, and vomiting."""]]).toDF("text")

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

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

val sentence_detector = new SentenceDetector()
      .setInputCols(Array("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_model = MedicalNerModel
      .pretrained("ner_clinical", "en", "clinical/models")
      .setInputCols(Array("sentence", "token", "embeddings"))
      .setOutputCol("clinical_ner")

val ner_model_converter = new NerConverterInternal()
      .setInputCols(Array("sentence", "token", "clinical_ner"))
      .setOutputCol("ner_chunk")

val chunk2doc = Chunk2Doc().setInputCols("ner_chunk").setOutputCol("ner_chunk_doc")

val sbert_embedder = BertSentenceEmbeddings
      .pretrained("sbiobert_base_cased_mli", "en","clinical/models")
      .setInputCols(Array("ner_chunk_doc"))
      .setOutputCol("sbert_embeddings")
    
val resolver = SentenceEntityResolverModel
      .pretrained("sbiobertresolve_umls_findings", "en", "clinical/models") 
      .setInputCols(Array("ner_chunk_doc", "sbert_embeddings")) 
      .setOutputCol("resolution")
      .setDistanceFunction("EUCLIDEAN")

val p_model = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, word_embeddings, ner_model, ner_model_converter, chunk2doc, sbert_embedder, resolver))
    
val data = Seq("A 28-year-old female with a history of gestational diabetes mellitus diagnosed eight years prior to presentation and subsequent type two diabetes mellitus (T2DM), one prior episode of HTG-induced pancreatitis three years prior to presentation, associated with an acute hepatitis, and obesity with a body mass index (BMI) of 33.5 kg/m2, presented with a one-week history of polyuria, polydipsia, poor appetite, and vomiting.").toDF("text")  

val res = p_model.fit(data).transform(data)

import nlu
nlu.load("en.resolve.umls.findings").predict("""A 28-year-old female with a history of gestational diabetes mellitus diagnosed eight years prior to presentation and subsequent type two diabetes mellitus (T2DM), one prior episode of HTG-induced pancreatitis three years prior to presentation, associated with an acute hepatitis, and obesity with a body mass index (BMI) of 33.5 kg/m2, presented with a one-week history of polyuria, polydipsia, poor appetite, and vomiting.""")

Results

|    | ner_chunk                             | cui_code   |
|---:|:--------------------------------------|:-----------|
|  0 | gestational diabetes mellitus         | C2183115   |
|  1 | subsequent type two diabetes mellitus | C3532488   |
|  2 | T2DM                                  | C3280267   |
|  3 | HTG-induced pancreatitis              | C4554179   |
|  4 | an acute hepatitis                    | C4750596   |
|  5 | obesity                               | C1963185   |
|  6 | a body mass index                     | C0578022   |
|  7 | polyuria                              | C3278312   |
|  8 | polydipsia                            | C3278316   |
|  9 | poor appetite                         | C0541799   |
| 10 | vomiting                              | C0042963   |

Model Information

Model Name: sbiobertresolve_umls_findings
Compatibility: Healthcare NLP 4.0.0+
License: Licensed
Edition: Official
Input Labels: [sentence_embeddings]
Output Labels: [name]
Language: en
Size: 2.2 GB
Case sensitive: false

References

Trained on 200K concepts from clinical findings.https://www.nlm.nih.gov/research/umls/index.html