RadReport AI is a sophisticated web application that automatically generates structured radiology reports from radiologist findings and uploaded images. It combines AI language processing with standardized templates to produce consistent, high-quality reports while saving valuable time for radiologists.

Study Type Selection: Support for Chest, Abdomen and Pelvis, or Full Body CT scans

Facility-Specific Templates: Customized technique sections based on facility equipment

AI-Enhanced Processing: Claude AI integration for natural language processing and image analysis

Structured Report Generation: Produces formatted reports with appropriate findings and impressions

Image Upload & Analysis: Optional image upload for AI-assisted analysis

Intelligent Pattern Matching: Maps findings to appropriate impressions using pattern database

Continuous Improvement: Logs unmatched findings for ongoing system enhancement

Python 3.8+

Supabase account

Anthropic Claude API key

Clone the repository

Create and activate virtual environment

Install dependencies

Create a .env file with your API keys

Run the application

Select Study Type: Choose between Full Body, Chest, or Abdomen and Pelvis

Choose Facility: Select the imaging facility where the scan was performed

Enter Findings: Input the radiologist's findings in the relevant sections

Upload Image (Optional): Add a CT scan image for additional AI analysis

Generate Report: Click the generate button to create a structured report

Review & Download: Review the generated report and download as needed

The system includes an admin panel (password protected) with the following capabilities:

Template Management: Modify facility-specific technique sections

Impression Pattern Database: Add and edit pattern-to-impression mappings

Unmatched Findings: Review findings that didn't match existing patterns and add new patterns based on them

Input Processing: The system processes the selected study type, facility, and entered findings

Template Selection: Appropriate templates are retrieved based on study type and facility

Finding Processing: Claude AI corrects grammar and formats findings

Category Matching: Each finding is matched to the appropriate category in the template

Impression Generation: The system matches findings to appropriate impressions using pattern matching

Image Analysis (Optional): If an image is provided, Claude analyzes it for additional findings

Report Assembly: All components are combined into a complete, structured report

DICOM image processing

Integration with hospital PACS/RIS systems

Enhanced AI analysis

Customizable report templates

Historical report comparison

Advanced analytics dashboard

Frontend: Streamlit

Database: Supabase

AI Processing: Anthropic Claude API

Deployment: Streamlit Cloud

Version Control: Git/GitHub

This project is proprietary and confidential. All rights reserved.

For questions or support, please contact ari@quantnexus.ai.