Description
Extract clinical entities from Romanian clinical texts. This model is trained using bert_base_cased embeddings.
Predicted Entities
Measurements, Form, Symptom, Route, Procedure, Disease_Syndrome_Disorder, Score, Drug_Ingredient, Pulse, Frequency, Date, Body_Part, Drug_Brand_Name, Time, Direction, Dosage, Medical_Device, Imaging_Technique, Test, Imaging_Findings, Imaging_Test, Test_Result, Weight, Clinical_Dept, Units
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "xx")\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
word_embeddings = BertEmbeddings.pretrained("bert_base_cased", "ro") \
.setInputCols(["sentence", "token"]) \
.setOutputCol("embeddings")
clinical_ner = MedicalNerModel.pretrained("ner_clinical_bert","ro","clinical/models")\
.setInputCols(["sentence","token","embeddings"])\
.setOutputCol("ner")
ner_converter = NerConverter()\
.setInputCols(["sentence","token","ner"])\
.setOutputCol("ner_chunk")
nlpPipeline = Pipeline(stages=[
document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
clinical_ner,
ner_converter])
sample_text = """ Solicitare: Angio CT cardio-toracic Dg. de trimitere Atrezie de valva pulmonara. Hipoplazie VS. Atrezie VAV stang. Anastomoza Glenn. Sp. Tromboza la nivelul anastomozei. Trimis de: Sectia Clinica Cardiologie (dr. Sue T.) Procedura Aparat GE Revolution HD. Branula albastra montata la nivelul membrului superior drept. Scout. Se administreaza 30 ml Iomeron 350 cu flux 2.2 ml/s, urmate de 20 ml ser fiziologic cu acelasi flux. Se efectueaza o examinare angio-CT cardiotoracica cu achizitii secventiale prospective la o frecventa cardiaca medie de 100/min."""
data = spark.createDataFrame([[sample_text]]).toDF("text")
result = nlpPipeline.fit(data).transform(data)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "xx")
.setInputCols(Array("document"))
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols(Array("sentence"))
.setOutputCol("token")
val word_embeddings = BertEmbeddings.pretrained("bert_base_cased", "ro")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val clinical_ner = MedicalNerModel.pretrained("ner_clinical_bert", "ro", "clinical/models")
.setInputCols(Array("sentence", "token", "embeddings"))
.setOutputCol("ner")
val ner_converter = new NerConverter()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("ner_chunk")
val pipeline = new Pipeline().setStages(Array(document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
clinical_ner,
ner_converter))
val data = Seq("""Solicitare: Angio CT cardio-toracic Dg. de trimitere Atrezie de valva pulmonara. Hipoplazie VS. Atrezie VAV stang. Anastomoza Glenn. Sp. Tromboza la nivelul anastomozei. Trimis de: Sectia Clinica Cardiologie (dr. Sue T.) Procedura Aparat GE Revolution HD. Branula albastra montata la nivelul membrului superior drept. Scout. Se administreaza 30 ml Iomeron 350 cu flux 2.2 ml/s, urmate de 20 ml ser fiziologic cu acelasi flux. Se efectueaza o examinare angio-CT cardiotoracica cu achizitii secventiale prospective la o frecventa cardiaca medie de 100/min.""").toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("ro.embed.clinical.bert.base_cased").predict(""" Solicitare: Angio CT cardio-toracic Dg. de trimitere Atrezie de valva pulmonara. Hipoplazie VS. Atrezie VAV stang. Anastomoza Glenn. Sp. Tromboza la nivelul anastomozei. Trimis de: Sectia Clinica Cardiologie (dr. Sue T.) Procedura Aparat GE Revolution HD. Branula albastra montata la nivelul membrului superior drept. Scout. Se administreaza 30 ml Iomeron 350 cu flux 2.2 ml/s, urmate de 20 ml ser fiziologic cu acelasi flux. Se efectueaza o examinare angio-CT cardiotoracica cu achizitii secventiale prospective la o frecventa cardiaca medie de 100/min.""")
Results
+--------------------------+-------------------------+
|chunks |entities |
+--------------------------+-------------------------+
|Angio CT cardio-toracic |Imaging_Test |
|Atrezie |Disease_Syndrome_Disorder|
|valva pulmonara |Body_Part |
|Hipoplazie |Disease_Syndrome_Disorder|
|VS |Body_Part |
|Atrezie |Disease_Syndrome_Disorder|
|VAV stang |Body_Part |
|Anastomoza Glenn |Disease_Syndrome_Disorder|
|Tromboza |Disease_Syndrome_Disorder|
|Sectia Clinica Cardiologie|Clinical_Dept |
|GE Revolution HD |Medical_Device |
|Branula albastra |Medical_Device |
|membrului superior drept |Body_Part |
|Scout |Body_Part |
|30 ml |Dosage |
|Iomeron 350 |Drug_Ingredient |
|2.2 ml/s |Dosage |
|20 ml |Dosage |
|ser fiziologic |Drug_Ingredient |
|angio-CT |Imaging_Test |
+--------------------------+-------------------------+
Model Information
| Model Name: | ner_clinical_bert |
| Compatibility: | Healthcare NLP 4.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | ro |
| Size: | 16.3 MB |
Benchmarking
label precision recall f1-score support
Body_Part 0.91 0.93 0.92 679
Clinical_Dept 0.68 0.65 0.67 97
Date 0.99 0.99 0.99 87
Direction 0.66 0.76 0.70 50
Disease_Syndrome_Disorder 0.73 0.76 0.74 121
Dosage 0.78 1.00 0.87 38
Drug_Ingredient 0.90 0.94 0.92 48
Form 1.00 1.00 1.00 6
Imaging_Findings 0.86 0.82 0.84 201
Imaging_Technique 0.92 0.92 0.92 26
Imaging_Test 0.93 0.98 0.95 205
Measurements 0.71 0.69 0.70 214
Medical_Device 0.85 0.81 0.83 42
Pulse 0.82 1.00 0.90 9
Route 1.00 0.91 0.95 33
Score 1.00 0.98 0.99 41
Time 1.00 1.00 1.00 28
Units 0.60 0.93 0.73 88
Weight 0.82 1.00 0.90 9
micro-avg 0.84 0.87 0.86 2037
macro-avg 0.70 0.74 0.72 2037
weighted-avg 0.84 0.87 0.85 2037