Upload folder using huggingface_hub
Browse files- README.md +93 -0
- config.json +6 -0
- figures/architecture.png +0 -0
- figures/license_badge.png +0 -0
- figures/performance_chart.png +0 -0
- pytorch_model.bin +3 -0
README.md
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: transformers
|
| 4 |
+
---
|
| 5 |
+
# MedicalBioModel
|
| 6 |
+
<!-- markdownlint-disable first-line-h1 -->
|
| 7 |
+
<!-- markdownlint-disable html -->
|
| 8 |
+
<!-- markdownlint-disable no-duplicate-header -->
|
| 9 |
+
|
| 10 |
+
<div align="center">
|
| 11 |
+
<img src="figures/architecture.png" width="60%" alt="MedicalBioModel" />
|
| 12 |
+
</div>
|
| 13 |
+
<hr>
|
| 14 |
+
|
| 15 |
+
<div align="center" style="line-height: 1;">
|
| 16 |
+
<a href="LICENSE" style="margin: 2px;">
|
| 17 |
+
<img alt="License" src="figures/license_badge.png" style="display: inline-block; vertical-align: middle;"/>
|
| 18 |
+
</a>
|
| 19 |
+
</div>
|
| 20 |
+
|
| 21 |
+
## 1. Introduction
|
| 22 |
+
|
| 23 |
+
The MedicalBioModel is a specialized biomedical language model designed for clinical NLP tasks. This model has been fine-tuned on a large corpus of clinical notes, medical literature, and healthcare data to achieve state-of-the-art performance on various medical benchmarks.
|
| 24 |
+
|
| 25 |
+
<p align="center">
|
| 26 |
+
<img width="80%" src="figures/performance_chart.png">
|
| 27 |
+
</p>
|
| 28 |
+
|
| 29 |
+
The model excels in tasks such as disease diagnosis prediction, drug interaction detection, medical named entity recognition, and clinical summarization. It has been trained with attention to patient safety and adverse event detection.
|
| 30 |
+
|
| 31 |
+
## 2. Clinical Benchmark Results
|
| 32 |
+
|
| 33 |
+
### Comprehensive Clinical Benchmark Results
|
| 34 |
+
|
| 35 |
+
<div align="center">
|
| 36 |
+
|
| 37 |
+
| | Benchmark | BioGPT | ClinicalBERT | PubMedBERT | MedicalBioModel |
|
| 38 |
+
|---|---|---|---|---|---|
|
| 39 |
+
| **Core Clinical Tasks** | Disease Diagnosis | 0.72 | 0.75 | 0.78 | 0.775 |
|
| 40 |
+
| | Drug Interaction | 0.65 | 0.68 | 0.70 | 0.773 |
|
| 41 |
+
| | Medical NER | 0.80 | 0.82 | 0.85 | 0.893 |
|
| 42 |
+
| **Clinical Understanding** | Clinical Notes | 0.70 | 0.73 | 0.76 | 0.825 |
|
| 43 |
+
| | Treatment Prediction | 0.62 | 0.65 | 0.68 | 0.699 |
|
| 44 |
+
| | Symptom Extraction | 0.68 | 0.71 | 0.74 | 0.881 |
|
| 45 |
+
| | Lab Result Interpretation | 0.72 | 0.74 | 0.77 | 0.800 |
|
| 46 |
+
| **Radiology & Imaging** | Radiology Report | 0.66 | 0.69 | 0.72 | 0.757 |
|
| 47 |
+
| | Patient Risk | 0.60 | 0.63 | 0.66 | 0.717 |
|
| 48 |
+
| **Clinical QA & Summary** | Medical QA | 0.68 | 0.71 | 0.74 | 0.817 |
|
| 49 |
+
| | Clinical Summary | 0.71 | 0.74 | 0.77 | 0.841 |
|
| 50 |
+
| **Safety & Compliance** | Adverse Event | 0.75 | 0.78 | 0.81 | 0.764 |
|
| 51 |
+
| | ICD Coding | 0.58 | 0.61 | 0.64 | 0.717 |
|
| 52 |
+
| | Clinical Trial Matching | 0.55 | 0.58 | 0.61 | 0.700 |
|
| 53 |
+
| | Patient Safety | 0.78 | 0.80 | 0.83 | 0.806 |
|
| 54 |
+
|
| 55 |
+
</div>
|
| 56 |
+
|
| 57 |
+
### Overall Clinical Performance Summary
|
| 58 |
+
The MedicalBioModel demonstrates strong performance across all evaluated clinical benchmark categories, with particularly notable results in patient safety and disease diagnosis tasks.
|
| 59 |
+
|
| 60 |
+
## 3. API Access & Demo
|
| 61 |
+
We offer a clinical NLP demo and API for you to interact with MedicalBioModel. Please check our official website for more details.
|
| 62 |
+
|
| 63 |
+
## 4. How to Run Locally
|
| 64 |
+
|
| 65 |
+
Please refer to our code repository for more information about running MedicalBioModel locally.
|
| 66 |
+
|
| 67 |
+
### System Prompt
|
| 68 |
+
We recommend using the following system prompt:
|
| 69 |
+
```
|
| 70 |
+
You are MedicalBioModel, a specialized medical AI assistant trained on clinical data.
|
| 71 |
+
Always prioritize patient safety and provide evidence-based information.
|
| 72 |
+
Today is {current date}.
|
| 73 |
+
```
|
| 74 |
+
|
| 75 |
+
### Temperature
|
| 76 |
+
For clinical applications, we recommend setting temperature to 0.3 for more consistent and reliable outputs.
|
| 77 |
+
|
| 78 |
+
### Clinical Note Processing
|
| 79 |
+
For processing clinical notes, please follow this template:
|
| 80 |
+
```
|
| 81 |
+
clinical_note_template = \
|
| 82 |
+
"""[Patient ID]: {patient_id}
|
| 83 |
+
[Clinical Note Begin]
|
| 84 |
+
{clinical_note_content}
|
| 85 |
+
[Clinical Note End]
|
| 86 |
+
{query}"""
|
| 87 |
+
```
|
| 88 |
+
|
| 89 |
+
## 5. License
|
| 90 |
+
This code repository is licensed under the [MIT License](LICENSE). The use of MedicalBioModel is subject to the [MIT License](LICENSE) and applicable healthcare data regulations.
|
| 91 |
+
|
| 92 |
+
## 6. Contact
|
| 93 |
+
If you have any questions, please raise an issue on our GitHub repository or contact us at clinical-ai@medicalbiomodel.org.
|
config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "biobert",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"BioBertForSequenceClassification"
|
| 5 |
+
]
|
| 6 |
+
}
|
figures/architecture.png
ADDED
|
figures/license_badge.png
ADDED
|
figures/performance_chart.png
ADDED
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e708defb1fb584bdc6a99c8581469f57acfe5a406c14d117354598e849523c9
|
| 3 |
+
size 228
|