Description
This Generic Classifier model is intended for detecting substance use in clinical notes and trained by using GenericClassifierApproach annotator. Present: if the patient was a current consumer of substance or the patient was a consumer in the past and had quit or if the patient had never consumed substance. None: if there was no related text.
Predicted Entities
Present, None
Live Demo Open in Colab Copy S3 URI
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_embeddings = BertSentenceEmbeddings.pretrained("sbiobert_base_cased_mli", 'en','clinical/models')\
.setInputCols(["document"])\
.setOutputCol("sentence_embeddings")
features_asm = FeaturesAssembler()\
.setInputCols(["sentence_embeddings"])\
.setOutputCol("features")
generic_classifier = GenericClassifierModel.pretrained("genericclassifier_sdoh_substance_usage_binary_sbiobert_cased_mli", 'en', 'clinical/models')\
.setInputCols(["features"])\
.setOutputCol("class")
pipeline = Pipeline(stages=[
document_assembler,
sentence_embeddings,
features_asm,
generic_classifier
])
text_list = ["Lives in apartment with 16-year-old daughter. Denies EtOH use currently although reports occasional use in past. Utox on admission positive for opiate (on as rx) as well as cocaine. 4-6 cigarettes a day on and off for 10 years. Denies h/o illicit drug use besides marijuana although admitted to cocaine use after being found to have urine positive for cocaine.",
"The patient quit smoking approximately two years ago with an approximately a 40 pack year history, mostly cigar use. He also reports 'heavy alcohol use', quit 15 months ago."]
df = spark.createDataFrame(text_list, StringType()).toDF("text")
result = pipeline.fit(df).transform(df)
result.select("text", "class.result").show(truncate=100)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence_embeddings = BertSentenceEmbeddings.pretrained("sbiobert_base_cased_mli", "en", "clinical/models")
.setInputCols("document")
.setOutputCol("sentence_embeddings")
val features_asm = new FeaturesAssembler()
.setInputCols("sentence_embeddings")
.setOutputCol("features")
val generic_classifier = GenericClassifierModel.pretrained("genericclassifier_sdoh_substance_usage_binary_sbiobert_cased_mli", "en", "clinical/models")
.setInputCols("features")
.setOutputCol("class")
val pipeline = new PipelineModel().setStages(Array(
document_assembler,
sentence_embeddings,
features_asm,
generic_classifier))
val data = Seq("The patient quit smoking approximately two years ago with an approximately a 40 pack year history, mostly cigar use. He also reports 'heavy alcohol use', quit 15 months ago.").toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("en.classify.generic.sdoh_substance_binary_sbiobert_cased").predict("""Lives in apartment with 16-year-old daughter. Denies EtOH use currently although reports occasional use in past. Utox on admission positive for opiate (on as rx) as well as cocaine. 4-6 cigarettes a day on and off for 10 years. Denies h/o illicit drug use besides marijuana although admitted to cocaine use after being found to have urine positive for cocaine.""")
Results
+----------------------------------------------------------------------------------------------------+---------+
| text| result|
+----------------------------------------------------------------------------------------------------+---------+
|Lives in apartment with 16-year-old daughter. Denies EtOH use currently although reports occasion...|[Present]|
|The patient quit smoking approximately two years ago with an approximately a 40 pack year history...| [None]|
+----------------------------------------------------------------------------------------------------+---------+
Model Information
| Model Name: | genericclassifier_sdoh_substance_usage_binary_sbiobert_cased_mli |
| Compatibility: | Healthcare NLP 4.2.4+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [features] |
| Output Labels: | [prediction] |
| Language: | en |
| Size: | 3.4 MB |
Benchmarking
label precision recall f1-score support
None 0.91 0.83 0.87 898
Present 0.76 0.87 0.81 540
accuracy - - 0.85 1438
macro-avg 0.83 0.85 0.84 1438
weighted-avg 0.85 0.85 0.85 1438