DIANA

DIANA provides DICOM imaging informatics services for high-resolution image archival, cohort discovery, radiation dose monitoring, PHI anonymization, and integration with PACS and clinical AI pipelines, implemented in Python.

Key Features:

• DICOM data handling and image indexing: Handles DICOM data and performs efficient image indexing for imaging studies.
• Interoperability with PACS: Acts as a lightweight interface to hospital Picture Archiving and Communications Systems (PACS) to enable programmatic access to imaging data.
• RESTful endpoints: Exposes RESTful endpoints for scripted access and automation.
• User access control via FOSS integration: Integrates free and open-source software components to provide user access control.
• AI integration: Supports clinical AI pipelines including bone age estimation and intra-cranial hemorrhage (ICH) detection, orchestrating workflows from study discovery post-acquisition to delivering online notifications.
• Performance efficiency: Demonstrates reduced latencies for clinical AI tasks with mean latencies of 9.04 ± 3.83 minutes (bone age estimation) and 20.17 ± 10.16 minutes (ICH detection), versus clinician times of 51.52 ± 58.9 minutes and 65.62 ± 110.39 minutes respectively.
• Data retrieval and anonymization: Retrieves and anonymizes protected health information (PHI) for large-scale imaging research with mean per-image retrieval times of 1.12 ± 0.50 seconds for x-ray studies and 0.08 ± 0.01 seconds for computed tomography studies.
• Scalability and flexible integration: Scales and integrates into hospital infrastructure to support large data retrieval and clinical integration of AI models.

Scientific Applications:

• Retrospective PACS data retrieval and cohort discovery: Enables large-scale retrieval and anonymization of PACS imaging data for research and cohort construction.
• Prospective clinical AI deployment: Orchestrates AI pipelines in clinical workflows, exemplified by bone age estimation and intra-cranial hemorrhage detection with online notification delivery.
• Radiation dose monitoring and archival: Supports high-resolution image archival and radiation dose monitoring from DICOM sources.
• Large-scale imaging dataset curation: Facilitates curation of imaging datasets for data-driven medicine and AI model development.

Methodology:

Performs DICOM data handling, efficient image indexing, exposes RESTful endpoints, performs PHI anonymization, integrates FOSS components for access control, and orchestrates study discovery post-acquisition with AI workflow execution and notification delivery.