Description
This model maps clinical entities and concepts (like drugs/ingredients) to National Drug Codes using sbiobert_base_cased_mli Sentence Bert Embeddings. It also returns package options and alternative drugs in the all_k_aux_label column.
Predicted Entities
NDC codes, package options
Live Demo Open in Colab Copy S3 URI
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentenceDetectorDL = 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("word_embeddings")
ner = MedicalNerModel.pretrained("ner_posology_greedy", "en", "clinical/models")\
.setInputCols(["sentence", "token", "word_embeddings"])\
.setOutputCol("ner")\
ner_converter = NerConverterInternal()\
.setInputCols(["sentence", "token", "ner"])\
.setOutputCol("ner_chunk")\
.setWhiteList(["DRUG"])
c2doc = Chunk2Doc()\
.setInputCols("ner_chunk")\
.setOutputCol("ner_chunk_doc")
sbert_embedder = BertSentenceEmbeddings.pretrained("sbiobert_base_cased_mli", "en", "clinical/models")\
.setInputCols(["ner_chunk_doc"])\
.setOutputCol("sentence_embeddings")\
.setCaseSensitive(False)
ndc_resolver = SentenceEntityResolverModel.pretrained("sbiobertresolve_ndc", "en", "clinical/models")\ \
.setInputCols(["sentence_embeddings"]) \
.setOutputCol("ndc_code")\
.setDistanceFunction("EUCLIDEAN")
resolver_pipeline = Pipeline(stages = [document_assembler,
sentenceDetectorDL,
tokenizer,
word_embeddings,
ner,
ner_converter,
c2doc,
sbert_embedder,
ndc_resolver])
text= "The patient was prescribed Amlodopine Vallarta 10-320mg, Eviplera. aspirin 81 mg and metformin 500 mg. The other patient is given Lescol 40 MG and Everolimus 1.5 mg tablet."
data = spark.createDataFrame([[text]]).toDF("text")
result = resolver_pipeline.fit(data).transform(data)
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_posology_greedy", "en", "clinical/models")
.setInputCols(Array("sentence","token","embeddings"))
.setOutputCol("ner")
val ner_converter = new NerConverter()
.setInputCols(Array("sentence","token","ner"))
.setOutputCol("ner_chunk")
.setWhiteList("DRUG")
val chunk2doc = new Chunk2Doc()
.setInputCols("ner_chunk")
.setOutputCol("ner_chunk_doc")
val sbert_embedder = BertSentenceEmbeddings.pretrained("sbiobert_base_cased_mli","en","clinical/models")
.setInputCols("ner_chunk_doc")
.setOutputCol("sbert_embeddings")
val ndc_resolver = SentenceEntityResolverModel.pretrained("sbiobertresolve_ndc", "en", "clinical/models")
.setInputCols("sbert_embeddings")
.setOutputCol("ndc_code")
.setDistanceFunction("EUCLIDEAN")
val pipeline = new Pipeline().setStages(Array(document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
ner,
ner_converter,
chunk2doc,
sbert_embedder,
ndc_resolver))
val data = Seq("The patient was prescribed Amlodopine Vallarta 10-320mg, Eviplera. aspirin 81 mg and metformin 500 mg. The other patient is given Lescol 40 MG and Everolimus 1.5 mg tablet.").toDS().toDF("text")
val result = pipeline.fit(data).transform(data)
Results
| | ner_chunk | entity | ndc_code | aux_list |
|---:|:-----------------------------|:---------|:-----------|:----------------------------------------------------------------------------------------------------|
| 0 | Amlodopine Vallarta 10-320mg | DRUG | 72483-0100 | '{'packages': "['1 BOTTLE in 1 BOX (72483-100-04) / 120 mL in 1 BOTTLE\']", \'alternatives\': [ ...|
| 1 | aspirin 81 mg | DRUG | 41250-0780 | '{'packages': "['1 BOTTLE, PLASTIC in 1 PACKAGE (41250-780-01) / 120 TABLET, DELAYED RELEASE in ...|
| 2 | metformin 500 mg | DRUG | 62207-0491 | '{'packages': "['5000 TABLET in 1 POUCH (62207-491-31)\', \'25000 TABLET in 1 CARTON (62207-491- ...|
| 3 | Lescol 40 MG | DRUG | 0713-0862 | '{'packages': "['30 TABLET, FILM COATED in 1 BOTTLE, PLASTIC (0713-0862-30)\']", \'alternatives\ ...|
| 4 | Everolimus 1.5 mg tablet | DRUG | 0054-0604 | '{'packages': "['60 TABLET in 1 BOTTLE (0054-0604-21)\']", \'alternatives\': [\'67877-721\', \'4 ...|
Model Information
| Model Name: | sbiobertresolve_ndc |
| Compatibility: | Healthcare NLP 5.4.1+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence_embeddings] |
| Output Labels: | [ndc_code] |
| Language: | en |
| Size: | 724.6 MB |
| Case sensitive: | false |
References
It is trained on U.S. FDA 2024-NDC Codes dataset