Description
Pretrained named entity recognition deep learning model for clinical terminology. The SparkNLP deep learning model (NerDL) is inspired by a former state of the art model for NER: Chiu & Nicols, Named Entity Recognition with Bidirectional LSTM-CNN.
Definitions of Predicted Entities:
Age: All mention of ages, past or present, related to the patient or with anybody else.Temperature: All mentions that refer to body temperature.O2_Saturation: Systemic arterial, venous or peripheral oxygen saturation measurements.Procedure: All mentions of invasive medical or surgical procedures or treatments.Symptom: All the symptoms mentioned in the document, of a patient or someone else.Substance: All mentions of substance use related to the patient or someone else (recreational drugs, illicit drugs).Route: Drug and medication administration routes available described by FDA.Blood_Pressure: Systemic blood pressure, mean arterial pressure, systolic and/or diastolic are extracted.Allergen: Allergen related extractions mentioned in the document.Gender: Gender-specific nouns and pronouns.Pulse: Peripheral heart rate, without advanced information like measurement location.Modifier: Terms that modify the symptoms, diseases or risk factors. If a modifier is included in ICD-10 name of a specific disease, the respective modifier is not extracted separately.Frequency: Frequency of administration for a dose prescribed.Weight: All mentions related to a patients weight.Dosage: Quantity prescribed by the physician for an active ingredient; measurement units are available described by FDA.Respiration: Number of breaths per minute.
Predicted Entities
Diagnosis, Procedure_Name, Lab_Result, Procedure, Procedure_Findings, O2_Saturation, Procedure_incident_description, Dosage, Causative_Agents_(Virus_and_Bacteria), Name, Cause_of_death, Substance_Name, Weight, Symptom_Name, Maybe, Modifier, Blood_Pressure, Frequency, Gender, Drug_incident_description, Age, Drug_Name, Temperature, Section_Name, Route, Negation, Negated, Allergenic_substance, Lab_Name, Respiratory_Rate.
Live Demo Open in Colab Copy S3 URI
How to use
Use as part of an nlp pipeline with the following stages: DocumentAssembler, SentenceDetector, Tokenizer, WordEmbeddingsModel, NerDLModel. Add the NerConverter to the end of the pipeline to convert entity tokens into full entity chunks.
...
embeddings_clinical = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")
clinical_ner = NerDLModel.pretrained("ner_jsl", "en", "clinical/models") \
.setInputCols(["sentence", "token", "embeddings"]) \
.setOutputCol("ner")
...
nlpPipeline = Pipeline(stages=[document_assembler, sentence_detector, tokenizer, embeddings_clinical, clinical_ner, ner_converter])
model = nlpPipeline.fit(spark.createDataFrame([[""]]).toDF("text"))
results = model.transform(spark.createDataFrame([["The patient is a 21-day-old Caucasian male here for 2 days of congestion - mom has been suctioning yellow discharge from the patient's nares, plus she has noticed some mild problems with his breathing while feeding (but negative for any perioral cyanosis or retractions). One day ago, mom also noticed a tactile temperature and gave the patient Tylenol. Baby also has had some decreased p.o. intake. His normal breast-feeding is down from 20 minutes q.2h. to 5 to 10 minutes secondary to his respiratory congestion. He sleeps well, but has been more tired and has been fussy over the past 2 days. The parents noticed no improvement with albuterol treatments given in the ER. His urine output has also decreased; normally he has 8 to 10 wet and 5 dirty diapers per 24 hours, now he has down to 4 wet diapers per 24 hours. Mom denies any diarrhea. His bowel movements are yellow colored and soft in nature."]], ["text"]))
...
val embeddings_clinical = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val ner = NerDLModel.pretrained("ner_jsl", "en", "clinical/models")
.setInputCols("sentence", "token", "embeddings")
.setOutputCol("ner")
...
val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, embeddings_clinical, ner, ner_converter))
val data = Seq("The patient is a 21-day-old Caucasian male here for 2 days of congestion - mom has been suctioning yellow discharge from the patient's nares, plus she has noticed some mild problems with his breathing while feeding (but negative for any perioral cyanosis or retractions). One day ago, mom also noticed a tactile temperature and gave the patient Tylenol. Baby also has had some decreased p.o. intake. His normal breast-feeding is down from 20 minutes q.2h. to 5 to 10 minutes secondary to his respiratory congestion. He sleeps well, but has been more tired and has been fussy over the past 2 days. The parents noticed no improvement with albuterol treatments given in the ER. His urine output has also decreased; normally he has 8 to 10 wet and 5 dirty diapers per 24 hours, now he has down to 4 wet diapers per 24 hours. Mom denies any diarrhea. His bowel movements are yellow colored and soft in nature.").toDF("text")
val result = pipeline.fit(data).transform(data)
Results
The output is a dataframe with a sentence per row and a "ner" column containing all of the entity labels in the sentence, entity character indices, and other metadata. To get only the tokens and entity labels, without the metadata, select "token.result" and "ner.result" from your output dataframe or add the "Finisher" to the end of your pipeline.
+---------------------------+------------+
|chunk |ner |
+---------------------------+------------+
|21-day-old |Age |
|male |Gender |
|congestion |Symptom_Name|
|mom |Gender |
|suctioning yellow discharge|Symptom_Name|
|she |Gender |
|mild |Modifier |
|problems with his breathing|Symptom_Name|
|negative |Negated |
|perioral cyanosis |Symptom_Name|
|retractions |Symptom_Name|
|mom |Gender |
|Tylenol |Drug_Name |
|His |Gender |
|his |Gender |
|respiratory congestion |Symptom_Name|
|He |Gender |
|tired |Symptom_Name|
|fussy |Symptom_Name|
|albuterol |Drug_Name |
+---------------------------+------------+
Model Information
| Model Name: | ner_jsl_en_2.4.2_2.4 |
| Type: | ner |
| Compatibility: | Spark NLP 2.4.2 |
| Edition: | Official |
| License: | Licensed |
| Input Labels: | [sentence,token, embeddings] |
| Output Labels: | [ner] |
| Language: | [en] |
| Case sensitive: | false |
Data Source
Trained on data gathered and manually annotated by John Snow Labs. https://www.johnsnowlabs.com/data/
Benchmarking
label tp fp fn prec rec f1
B-Pulse_Rate 77 39 12 0.663793 0.865169 0.75122
I-Diagnosis 2134 1139 1329 0.652001 0.616229 0.63361
I-Procedure_Name 2335 1329 956 0.637282 0.709511 0.671459
B-Lab_Result 601 182 94 0.767561 0.864748 0.813261
B-Procedure 1 0 5 1 0.166667 0.285714
B-Procedure_Findings 2 13 72 0.133333 0.027027 0.044944
B-O2_Saturation 1 3 4 0.25 0.2 0.222222
B-Dosage 477 197 68 0.707715 0.875229 0.782609
I-Causative_Agents_(Virus_and_Bacteria) 12 2 7 0.857143 0.631579 0.727273
B-Name 562 268 554 0.677108 0.503584 0.577595
I-Cause_of_death 9 5 11 0.642857 0.45 0.529412
I-Substance_Name 24 34 54 0.413793 0.307692 0.352941
I-Name 716 377 710 0.655078 0.502104 0.56848
B-Cause_of_death 9 6 8 0.6 0.529412 0.5625
B-Weight 52 22 9 0.702703 0.852459 0.77037
B-Symptom_Name 4364 1916 1652 0.694904 0.725399 0.709824
I-Maybe 27 51 61 0.346154 0.306818 0.325301
I-Symptom_Name 2073 1492 2348 0.581487 0.468898 0.519159
B-Modifier 1573 890 768 0.638652 0.671935 0.654871
B-Blood_Pressure 76 19 13 0.8 0.853933 0.826087
B-Frequency 308 134 77 0.696833 0.8 0.744861
I-Gender 26 31 28 0.45614 0.481482 0.468468
I-Drug_incident_description 4 10 57 0.285714 0.065574 0.106667
B-Drug_incident_description 2 5 23 0.285714 0.08 0.125
I-Age 5 0 9 1 0.357143 0.526316
B-Drug_Name 1741 490 290 0.780368 0.857213 0.816987
B-Substance_Name 148 41 48 0.783069 0.755102 0.768831
B-Temperature 56 23 13 0.708861 0.811594 0.756757
I-Procedure 1 0 7 1 0.125 0.222222
B-Section_Name 2711 317 166 0.89531 0.942301 0.918205
I-Route 119 110 189 0.519651 0.386364 0.443203
B-Maybe 143 80 127 0.641256 0.52963 0.580122
B-Gender 5166 709 58 0.879319 0.988897 0.930895
I-Dosage 434 196 87 0.688889 0.833013 0.754127
B-Causative_Agents_(Virus_and_Bacteria) 19 3 8 0.863636 0.703704 0.77551
I-Frequency 275 134 191 0.672372 0.590129 0.628571
B-Age 357 27 16 0.929688 0.957105 0.943197
I-Lab_Result 45 78 152 0.365854 0.228426 0.28125
B-Negation 99 38 38 0.722628 0.722628 0.722628
B-Diagnosis 2786 1342 913 0.674903 0.753177 0.711895
I-Section_Name 3885 1353 179 0.741695 0.955955 0.835304
B-Route 421 217 166 0.659875 0.717206 0.687347
I-Negation 11 30 24 0.268293 0.314286 0.289474
B-Procedure_Name 1490 811 522 0.647545 0.740557 0.690934
B-Negated 1490 332 215 0.817783 0.8739 0.844911
I-Allergenic_substance 1 0 12 1 0.0769231 0.142857
I-Negated 89 132 146 0.402715 0.378723 0.390351
I-Procedure_Findings 2 31 283 0.060606 0.0070175 0.012570
B-Allergenic_substance 72 29 24 0.712871 0.75 0.730965
I-Weight 47 35 16 0.573171 0.746032 0.648276
B-Lab_Name 804 290 122 0.734918 0.868251 0.79604
I-Modifier 99 73 422 0.575581 0.190019 0.285714
I-Temperature 1 0 14 1 0.066667 0.125
I-Drug_Name 362 284 261 0.560372 0.581059 0.570528
I-Lab_Name 284 194 127 0.594142 0.690998 0.63892
B-Respiratory_Rate 46 5 5 0.901961 0.901961 0.901961
Macro-average 38674 15571 13819 0.589085 0.515426 0.5498
Micro-average 38674 15571 13819 0.712951 0.736746 0.724653