|
|
--- |
|
|
license: apache-2.0 |
|
|
library_name: transformers |
|
|
--- |
|
|
# MedAssistPro |
|
|
<!-- markdownlint-disable first-line-h1 --> |
|
|
<!-- markdownlint-disable html --> |
|
|
<!-- markdownlint-disable no-duplicate-header --> |
|
|
|
|
|
<div align="center"> |
|
|
<img src="figures/fig1.png" width="60%" alt="MedAssistPro" /> |
|
|
</div> |
|
|
<hr> |
|
|
|
|
|
<div align="center" style="line-height: 1;"> |
|
|
<a href="LICENSE" style="margin: 2px;"> |
|
|
<img alt="License" src="figures/fig2.png" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
</div> |
|
|
|
|
|
## 1. Introduction |
|
|
|
|
|
MedAssistPro represents a breakthrough in medical AI assistants. This latest version incorporates advanced clinical knowledge integration and has been extensively trained on de-identified medical records, clinical guidelines, and peer-reviewed literature. The model excels in understanding complex medical terminology and providing evidence-based clinical decision support. |
|
|
|
|
|
<p align="center"> |
|
|
<img width="80%" src="figures/fig3.png"> |
|
|
</p> |
|
|
|
|
|
Compared to our previous release, MedAssistPro demonstrates remarkable improvements in clinical reasoning capabilities. In the MedQA-USMLE benchmark, accuracy improved from 65% to 82.3%. This enhancement stems from our novel medical knowledge distillation technique that ensures both accuracy and interpretability in clinical contexts. |
|
|
|
|
|
Beyond diagnostic support, MedAssistPro also features improved drug interaction detection, enhanced ICD-10 coding accuracy, and better clinical note summarization capabilities. |
|
|
|
|
|
## 2. Evaluation Results |
|
|
|
|
|
### Comprehensive Benchmark Results |
|
|
|
|
|
<div align="center"> |
|
|
|
|
|
| | Benchmark | ClinicalBERT | BioBERT | PubMedGPT | MedAssistPro | |
|
|
|---|---|---|---|---|---| |
|
|
| **Diagnostic Tasks** | Diagnosis Accuracy | 0.712 | 0.735 | 0.751 | 0.711 | |
|
|
| | Drug Interaction | 0.689 | 0.701 | 0.723 | 0.774 | |
|
|
| | Medical QA | 0.756 | 0.772 | 0.789 | 0.884 | |
|
|
| **Clinical Documentation** | Clinical Notes | 0.671 | 0.685 | 0.702 | 0.658 | |
|
|
| | Radiology Report | 0.622 | 0.641 | 0.659 | 0.676 | |
|
|
| | Medical Summarization | 0.695 | 0.711 | 0.728 | 0.809 | |
|
|
| **Clinical Decision Support** | Patient Triage | 0.733 | 0.749 | 0.761 | 0.839 | |
|
|
| | Symptom Extraction | 0.788 | 0.801 | 0.815 | 0.766 | |
|
|
| | Treatment Recommendation | 0.651 | 0.668 | 0.682 | 0.728 | |
|
|
| **Coding & Classification**| Lab Result Interpretation | 0.702 | 0.718 | 0.731 | 0.804 | |
|
|
| | ICD Coding | 0.768 | 0.781 | 0.793 | 0.740 | |
|
|
| | Adverse Event Detection | 0.718 | 0.729 | 0.745 | 0.797 | |
|
|
|
|
|
</div> |
|
|
|
|
|
### Overall Performance Summary |
|
|
MedAssistPro achieves state-of-the-art performance across clinical NLP benchmarks, with particularly strong results in diagnostic accuracy and clinical reasoning tasks. |
|
|
|
|
|
## 3. Clinical Integration & API |
|
|
We provide a HIPAA-compliant API for healthcare institutions to integrate MedAssistPro. Contact our enterprise team for deployment options. |
|
|
|
|
|
## 4. How to Run Locally |
|
|
|
|
|
Please refer to our clinical deployment guide for information about running MedAssistPro in your healthcare environment. |
|
|
|
|
|
Key configuration recommendations: |
|
|
|
|
|
1. Enable clinical context mode for optimal performance. |
|
|
2. Configure confidence thresholds appropriate for your clinical workflow. |
|
|
|
|
|
### System Prompt |
|
|
We recommend the following clinical system prompt: |
|
|
``` |
|
|
You are MedAssistPro, a clinical decision support assistant. |
|
|
Today is {current date}. |
|
|
Always provide evidence-based recommendations with appropriate citations. |
|
|
``` |
|
|
|
|
|
### Temperature |
|
|
For clinical applications, we recommend setting temperature to 0.3 for consistent, deterministic outputs. |
|
|
|
|
|
### Integration with EHR Systems |
|
|
MedAssistPro supports FHIR R4 and HL7v2 message formats for seamless EHR integration. |
|
|
``` |
|
|
ehr_config = { |
|
|
"fhir_endpoint": "{fhir_server_url}", |
|
|
"auth_method": "oauth2", |
|
|
"patient_context": True |
|
|
} |
|
|
``` |
|
|
|
|
|
## 5. License |
|
|
This model is licensed under the [Apache 2.0 License](LICENSE). Usage in clinical settings requires compliance with local medical device regulations and appropriate clinical validation. |
|
|
|
|
|
## 6. Contact |
|
|
For enterprise inquiries and clinical partnership opportunities, contact us at enterprise@medassistpro.health. |
|
|
|