Description
This model is a modified version of Flan-T5 (LLM) based summarization model that is finetuned with clinical notes, encounters, critical care notes, discharge notes, reports, curated by John Snow Labs. This model is further optimized by augmenting the training methodology, and dataset. It can generate summaries from clinical notes up to 512 tokens given the input text (max 1024 tokens).
Predicted Entities
Live Demo Open in Colab Copy S3 URI
How to use
document = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
summarizer = MedicalSummarizer()\
.pretrained("summarizer_clinical_jsl", "en", "clinical/models")\
.setInputCols("document")\
.setOutputCol("summary")\
.setMaxTextLength(512)\
.setMaxNewTokens(512)
pipeline = Pipeline(stages=[document,summarizer])
text = """Patient with hypertension, syncope, and spinal stenosis - for recheck.
(Medical Transcription Sample Report)
SUBJECTIVE:
The patient is a 78-year-old female who returns for recheck. She has hypertension. She denies difficulty with chest pain, palpations, orthopnea, nocturnal dyspnea, or edema.
PAST MEDICAL HISTORY / SURGERY / HOSPITALIZATIONS:
Reviewed and unchanged from the dictation on 12/03/2003.
MEDICATIONS:
Atenolol 50 mg daily, Premarin 0.625 mg daily, calcium with vitamin D two to three pills daily, multivitamin daily, aspirin as needed, and TriViFlor 25 mg two pills daily. She also has Elocon cream 0.1% and Synalar cream 0.01% that she uses as needed for rash."""
data = spark.createDataFrame([[text]]).toDF("text")
result = pipeline.fit(data).transform(data)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val summarizer = MedicalSummarizer()
.pretrained("summarizer_clinical_jsl", "en", "clinical/models")
.setInputCols("document")
.setOutputCol("summary")
.setMaxTextLength(512)
.setMaxNewTokens(512)
val pipeline = new Pipeline().setStages(Array(document_assembler, summarizer))
val text = """Patient with hypertension, syncope, and spinal stenosis - for recheck.
(Medical Transcription Sample Report)
SUBJECTIVE:
The patient is a 78-year-old female who returns for recheck. She has hypertension. She denies difficulty with chest pain, palpations, orthopnea, nocturnal dyspnea, or edema.
PAST MEDICAL HISTORY / SURGERY / HOSPITALIZATIONS:
Reviewed and unchanged from the dictation on 12/03/2003.
MEDICATIONS:
Atenolol 50 mg daily, Premarin 0.625 mg daily, calcium with vitamin D two to three pills daily, multivitamin daily, aspirin as needed, and TriViFlor 25 mg two pills daily. She also has Elocon cream 0.1% and Synalar cream 0.01% that she uses as needed for rash."""
val data = Seq(text).toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
Results
A 78-year-old female with hypertension, syncope, and spinal stenosis returns for recheck. She denies chest pain, palpations, orthopnea, nocturnal dyspnea, or edema. She is on multiple medications and has Elocon cream and Synalar cream for rash.
Model Information
| Model Name: | summarizer_clinical_jsl |
| Compatibility: | Healthcare NLP 4.3.1+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 920.1 MB |
Benchmarkings
Benchmark on MtSamples Summarization Dataset :
| model_name | model_size | rouge | bleu | bertscore_precision | bertscore_recall: | bertscore_f1 |
|---|---|---|---|---|---|---|
| philschmid/flan-t5-base-samsum | 250M | 0.1919 | 0.1124 | 0.8409 | 0.8964 | 0.8678 |
| linydub/bart-large-samsum | 500M | 0.1586 | 0.0732 | 0.8747 | 0.8184 | 0.8456 |
| philschmid/bart-large-cnn-samsum | 500M | 0.2170 | 0.1299 | 0.8846 | 0.8436 | 0.8636 |
| transformersbook/pegasus-samsum | 500M | 0.1924 | 0.0965 | 0.8920 | 0.8149 | 0.8517 |
| summarizer_clinical_jsl | 250M | 0.4836 | 0.4188 | 0.9041 | 0.9374 | 0.9204 |
| summarizer_clinical_jsl_augmented | 250M | 0.5119 | 0.4545 | 0.9282 | 0.9526 | 0.9402 |
Benchmark on MIMIC Summarization Dataset :
| model_name | model_size | rouge | bleu | bertscore_precision | bertscore_recall: | bertscore_f1 |
|---|---|---|---|---|---|---|
| philschmid/flan-t5-base-samsum | 250M | 0.1910 | 0.1037 | 0.8708 | 0.9056 | 0.8879 |
| linydub/bart-large-samsum | 500M | 0.1252 | 0.0382 | 0.8933 | 0.8440 | 0.8679 |
| philschmid/bart-large-cnn-samsum | 500M | 0.1795 | 0.0889 | 0.9172 | 0.8978 | 0.9074 |
| transformersbook/pegasus-samsum | 570M | 0.1425 | 0.0582 | 0.9171 | 0.8682 | 0.8920 |
| summarizer_clinical_jsl | 250M | 0.395 | 0.2962 | 0.895 | 0.9316 | 0.913 |
| summarizer_clinical_jsl_augmented | 250M | 0.3964 | 0.307 | 0.9109 | 0.9452 | 0.9227 |
References
Trained on in-house curated dataset