Clinical Report Generator

This Hugging Face Space provides a clinical report generation service using a fine-tuned T5 model. The model has been trained on a dataset of clinical documents to generate professional, structured clinical reports from input notes.

Features

Generates professional clinical reports from input notes
Maintains medical terminology and context
Removes unwanted elements like URLs
Provides a user-friendly interface

How to Use

Enter your clinical notes in the input text box
Click "Submit" to generate the report
The generated report will appear in the output text box

Model Details

Base Model: T5-small
Fine-tuned on: Clinical documents and structured templates
Input Length: Up to 512 tokens
Output Length: Up to 256 tokens
Special Features:
- Beam search with 4 beams
- No repeat ngram size of 3
- Length penalty of 2.0
- Early stopping enabled
- URL filtering

Examples

The interface includes example inputs to demonstrate the model's capabilities. These examples cover common clinical scenarios such as:

Initial patient presentations
Follow-up visits
Treatment responses

Notes

The model is designed for generating clinical reports only
Input should be clear and contain relevant clinical information
Generated reports should be reviewed by healthcare professionals

SDK Version

sdk_version: 4.12.0