Sentence Entity Resolver for ICD-10-CM Codes (sbertresolve_icd10cm_augmented)

Description

This model maps clinical entities and concepts to ICD-10-CM codes using sbert_jsl_medium_uncased sentence bert embeddings. It also returns the official resolution text within the brackets inside the metadata. The model is augmented with synonyms, and previous augmentations are flexed according to cosine distances to unnormalized terms (ground truths).

Predicted Entities

ICD-10-CM Codes

Open in Colab Download 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")

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

ner_converter = NerConverterInternal()\
    .setInputCols(["sentence","token","ner"])\
    .setOutputCol("ner_chunk")\
    .setWhiteList(["PROBLEM"])

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

bert_embeddings = BertSentenceEmbeddings.pretrained("sbert_jsl_medium_uncased", "en", "clinical/models")\
    .setInputCols(["ner_chunk_doc"])\
    .setOutputCol("bert_embeddings")\
    .setCaseSensitive(False)

icd10_resolver = SentenceEntityResolverModel.pretrained("sbertresolve_icd10cm_augmented", "en", "clinical/models")\
    .setInputCols(["bert_embeddings"]) \
    .setOutputCol("resolution")\
    .setDistanceFunction("EUCLIDEAN")

nlpPipeline = Pipeline(stages=[document_assembler, 
                               sentence_detector, 
                               tokenizer, 
                               word_embeddings, 
                               clinical_ner, 
                               ner_converter, 
                               chunk2doc, 
                               bert_embeddings, 
                               icd10_resolver])

data_ner = 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, three years prior to presentation, associated with 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 = nlpPipeline.fit(data_ner).transform(data_ner)

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 clinical_ner = MedicalNerModel.pretrained("ner_clinical", "en", "clinical/models")
    .setInputCols(Array("sentence","token","embeddings"))
    .setOutputCol("ner")

val ner_converter = new NerConverterInternal()
    .setInputCols(Array("sentence","token","ner"))
    .setOutputCol("ner_chunk")
    .setWhiteList(Array("PROBLEM"))

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

val bert_embeddings = BertSentenceEmbeddings.pretrained("sbert_jsl_medium_uncased", "en", "clinical/models")
    .setInputCols("ner_chunk_doc")
    .setOutputCol("bert_embeddings")
    .setCaseSensitive(False)

val icd10_resolver = SentenceEntityResolverModel.pretrained("sbertresolve_icd10cm_augmented", "en", "clinical/models")
    .setInputCols("bert_embeddings")
    .setOutputCol("resolution")
    .setDistanceFunction("EUCLIDEAN")
    
val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, word_embeddings, clinical_ner, ner_converter, chunk2doc, sbert_embedder, icd10_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, three years prior to presentation, associated with 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 result = pipeline.fit(data).transform(data)

Results

+-------------------------------------+-------+----------+---------------------------------------------------------------------------+---------------------------------------------------------------------------+
|                            ner_chunk| entity|icd10_code|                                                                resolutions|                                                                  all_codes|
+-------------------------------------+-------+----------+---------------------------------------------------------------------------+---------------------------------------------------------------------------+
|        gestational diabetes mellitus|PROBLEM|     O24.4|[gestational diabetes mellitus [gestational diabetes mellitus], maternal...|     [O24.4, O24.41, O24.43, Z86.32, K86.8, P70.2, O24.434, E10.9, O24.430]|
|subsequent type two diabetes mellitus|PROBLEM|       E11|[type 2 diabetes mellitus [type 2 diabetes mellitus], type ii diabetes m...|[E11, E11.9, E10.9, E10, E13.9, Z83.3, L83, E11.8, E11.32, E10.8, Z86.39...|
|                      acute hepatitis|PROBLEM|     K72.0|[acute hepatitis [acute and subacute hepatic failure], acute hepatitis a...|[K72.0, B15, B17.2, B17.1, B16, B17.9, B18.8, B15.9, K75.2, K73.9, B17.1...|
|                              obesity|PROBLEM|     E66.9|[obesity [obesity, unspecified], upper body obesity [other obesity], chi...|                                  [E66.9, E66.8, P90, Q13.0, M79.4, Z86.39]|
|                    a body mass index|PROBLEM|     E66.9|[observation of body mass index [obesity, unspecified], finding of body ...|[E66.9, Z68.41, Z68, E66.8, Z68.45, Z68.4, Z68.1, Z68.2, R22.9, Z68.22, ...|
|                             polyuria|PROBLEM|       R35|[polyuria [polyuria], sialuria [other specified metabolic disorders], st...|[R35, E88.8, R30.0, N28.89, O04.8, R82.4, E74.8, R82.2, E73.9, R82.0, R3...|
|                           polydipsia|PROBLEM|     R63.1|[polydipsia [polydipsia], polyotia [accessory auricle], polysomia [conjo...|[R63.1, Q17.0, Q89.4, Q89.09, Q74.8, H53.8, H53.2, Q13.2, R63.8, E23.2, ...|
|                        poor appetite|PROBLEM|     R63.0|[poor appetite [anorexia], excessive appetite [polyphagia], poor feeding...|[R63.0, R63.2, P92.9, R45.81, Z55.8, R41.84, R41.3, Z74.8, R46.89, R45.8...|
|                             vomiting|PROBLEM|     R11.1|[vomiting [vomiting], vomiting bile [vomiting following gastrointestinal...|[R11.1, K91.0, K92.0, A08.39, R11, P92.0, P92.09, R11.12, R11.10, O21.9,...|
+-------------------------------------+-------+----------+---------------------------------------------------------------------------+---------------------------------------------------------------------------+

Model Information

Model Name:	sbertresolve_icd10cm_augmented
Compatibility:	Healthcare NLP 4.4.2+
License:	Licensed
Edition:	Official
Input Labels:	[bert_embeddings]
Output Labels:	[icd10cm_code]
Language:	en
Size:	938.1 MB
Case sensitive:	false

PREVIOUSSentence Entity Resolver for Billable ICD10-CM HCC Codes (sbertresolve_icd10cm_augmented_billable_hcc)

NEXTSentence Entity Resolver for Hierarchical Condition Categories (HCC) codes (Augmented)