README.md

metadata

title: Medical Transcription Notes
emoji: 📝
colorFrom: gray
colorTo: yellow
sdk: gradio
sdk_version: 5.35.0
app_file: app.py
pinned: false
license: apache-2.0
short_description: 'OnDevice medical transcription and note creation '

Medical Transcription & Note Generation

This Hugging Face Space provides an end-to-end solution for medical transcription and structured note generation from patient-doctor conversations.

Paper

https://www.medrxiv.org/content/10.1101/2025.07.01.25330679v1

Features

🎙️ Audio Recording & Transcription

• Multi-source Audio Input: Record directly via microphone or upload audio files
• Multiple Whisper Models: Choose from all available Whisper models (tiny to large-v3)
• GPU-Accelerated: Utilizes GPU for fast transcription when available
• Streaming Output: Real-time transcription display with editable text
• Audio Download: Save recorded conversations as WAV files

📋 Medical Note Generation

• Side-by-side View: Transcription and medical notes displayed together
• Multiple LLM Models: Select from various medical-focused language models
• Dual Note Types:
  • SOAP Notes: Structured Subjective, Objective, Assessment, Plan format
  • Full Structured Notes: Comprehensive 14-section clinical documentation
• Copy Functionality: Easy copying of transcription and generated notes

Note Types

SOAP Note Format

• S (Subjective): Patient's reported symptoms and history
• O (Objective): Clinical findings, vital signs, lab results
• A (Assessment): Diagnosis and clinical reasoning
• P (Plan): Treatment plan and management

Full Structured Medical Note

1. Presenting Illness
2. History of Presenting Illness
3. Past Medical History
4. Surgical History
5. Family History
6. Social History
7. Allergy History
8. Medication History
9. Dietary History
10. Review of Systems
11. Physical Exam Findings
12. Labs and Imaging
13. Assessment
14. Plan

Models Available

Whisper Models

• openai/whisper-tiny (Default - fastest)
• openai/whisper-base
• openai/whisper-small
• openai/whisper-medium
• openai/whisper-large-v2
• openai/whisper-large-v3

Medical LLM Models

• OnDeviceMedNotes/Struct_Med_Note_v01 (Default)
• microsoft/DialoGPT-medium
• meta-llama/Llama-2-7b-chat-hf
• mistralai/Mistral-7B-Instruct-v0.1

Usage Instructions

1. Record/Upload Audio: Use the microphone to record or upload an audio file
2. Select Whisper Model: Choose transcription model based on speed/accuracy needs
3. Transcribe: Click "Transcribe Audio" to convert speech to text
4. Edit Transcription: Review and edit the transcribed text if needed
5. Choose Note Type: Select between SOAP note or full structured note
6. Select LLM Model: Choose the medical language model
7. Generate Note: Click "Generate Medical Note" to create structured documentation
8. Copy/Download: Use copy buttons or download audio file as needed

Technical Details

• GPU Support: Automatic GPU utilization for both transcription and note generation
• Memory Optimization: Efficient model loading and inference
• Error Handling: Robust error handling for various input formats
• Audio Processing: Automatic resampling and format conversion

Privacy & Security

⚠️ Important: This application is for demonstration purposes. Do not upload actual patient data or PHI (Protected Health Information). Always ensure compliance with HIPAA and local privacy regulations when handling medical data.

License

This project is open source. Please ensure appropriate licensing for any models used in production environments.