A real-time DICOM relay and analysis system with AI integration, persistent history, and a live web dashboard for reviewing and flagging X-ray images.

• Asynchronous TCP DICOM server (storescp-compatible)
• Real-time queue processing with AI API integration
• Automatic DICOM-to-PNG conversion (OpenCV)
• Persistent processing history with SQLite storage
• WebSocket-powered live dashboard updates
• Manual review of processed items (normal/abnormal)
• Manual DICOM QueryRetrieve trigger with configurable time span
• Automatic QueryRetrieve of CR modality studies at configurable intervals
• Fully responsive PicoCSS dashboard with lightbox image previews
• Comprehensive statistics and performance metrics
• Configuration file support
• Patient age calculation from Romanian ID numbers
• Anatomic region identification and region-specific analysis
• Previous report comparison for longitudinal studies
• Notification system for positive findings via ntfy.sh
• Database backup and maintenance routines
• Logging with timestamps to both console and file
• FHIR integration for patient data and radiologist reports
• AI-powered analysis of radiologist reports for quality metrics
• Report checking functionality for free-text radiology reports

• Python 3.8+
• DICOM peer system for QueryRetrieve (pynetdicom-compatible)
• OpenAI API endpoint (can be local or remote)

pip install aiohttp pydicom pynetdicom opencv-python numpy

1. Clone the repository:

git clone https://github.com/cstroie/XRayVision.git
cd XRayVision

1. Update the configuration in xrayvision.cfg:

[openai]
OPENAI_URL_PRIMARY = http://127.0.0.1:8080/v1/chat/completions
OPENAI_URL_SECONDARY = http://127.0.0.1:11434/v1/chat/completions
OPENAI_API_KEY = sk-your-api-key
MODEL_NAME = medgemma-4b-it

[dicom]
AE_TITLE = XRAYVISION
AE_PORT = 4010
REMOTE_AE_TITLE = DICOM_SERVER
REMOTE_AE_IP = 192.168.1.1
REMOTE_AE_PORT = 104

1. Run the server:

export OPENAI_API_KEY="sk-your-api-key"
python xrayvision.py

Optional arguments:

• --keep-dicom - Do not delete .dcm files after conversion
• --load-dicom - Load existing .dcm files in queue
• --no-query - Do not query the DICOM server automatically
• --enable-ntfy - Enable ntfy.sh notifications
• --model - Model name to use for analysis
• --retrieval-method - DICOM retrieval method (C-MOVE or C-GET)
• --log-level - Set logging level (DEBUG, INFO, WARNING, ERROR)

1. Open the dashboard:

http://localhost:8000

• Live processing statistics (queue, current file, success, failure)
• Paginated processed files with thumbnails
• Real-time review functionality (normal/abnormal)
• Manual QueryRetrieve trigger (select time span: 1, 3, 6, 12, 24 hours)
• Lightbox image preview with reviewed and positive highlighting
• Comprehensive statistics page with charts and metrics
• Configuration display
• Search and filtering capabilities
• Patient details view
• Radiologist report integration and analysis
• Report checking page for free-text analysis

XRayVision uses a configuration file (xrayvision.cfg) for all settings. A default configuration is provided in the file, and you can override settings by creating a local.cfg file with your custom values.

Key configuration sections include:

• general - Database path, backup directory
• users - User credentials with roles (admin, user)
• dicom - DICOM server settings (AE title, port, remote server details)
• openai - OpenAI API endpoints and credentials
• dashboard - Dashboard port
• notifications - ntfy.sh notification URL
• processing - Processing options (page size, DICOM file handling, query settings)
• regions - Anatomic region identification rules (keywords for region detection)
• questions - Region-specific questions for AI analysis
• supported_regions - List of regions to process (enable/disable regions)

All events are logged to:

• xrayvision.log file
• Console output

Timestamps, info, warnings, and errors are all captured.

• / - Main dashboard
• /stats - Statistics page
• /about - About page
• /check - Report check page
• /ws - WebSocket for real-time updates
• / - Main dashboard
• /stats - Statistics page
• /about - About page
• /check - Report check page
• /ws - WebSocket for real-time updates
• /api/exams - Get exams with pagination and filtering
• /api/stats - Get statistics data
• /api/config - Get configuration parameters
• /api/regions - Get supported regions
• /api/patients - Get patients with pagination and filtering
• /api/patients/{cnp} - Get patient by CNP
• /api/exams/{uid} - Get exam by UID
• /api/radreview - Record radiologist review of an exam
• /api/requeue - Re-queue an exam for processing
• /api/dicomquery - Manually trigger DICOM QueryRetrieve
• /api/check - Analyze a free-text radiology report
• /api/getrad - Retrieve radiologist report from FHIR
• /api/spec - Get OpenAPI specification

• Enhanced user management with role-based access control
• More detailed audit logging
• Export functionality for reports
• Integration with more DICOM modalities
• Improved error handling and recovery mechanisms
• Enhanced statistics and reporting capabilities