OsiriXGPT: An Innovative AI Co-pilot Plug-In for Seamless Deployment of Generative AI Models in Scan-to-Scan Reporting Workflows
Revolutionizing Radiology: Seamless AI Integration for Enhanced Workflows
Generative AI (GenAI) holds immense potential to revolutionize radiology by alleviating reporting burdens, enhancing diagnostic workflows, and improving communication of complex radiological information. However, its adoption is hampered by the lack of seamless integration with existing medical imaging viewers. This study introduces OsiriXgrpc, an open-source API plug-in that bridges this gap, enabling real-time communication between OsiriX (a CE-marked and FDA-approved DICOM viewer) and AI-driven tools developed in various programming languages (e.g., Python). OsiriXgrpc offers a unified platform for users to query, interact with, and visualize AI-generated outputs directly within OsiriX, addressing a critical unmet clinical need in radiology.
Executive Impact Metrics
0
Overall Classification Accuracy
0
Latency (Speech-to-Text)
0
OCR Redaction Accuracy
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Integration & Workflow
Generative AI Models
Clinical Impact
The core innovation is OsiriXgrpc, an API plug-in facilitating real-time communication between OsiriX and AI tools. This bridges the gap between powerful GenAI models and established medical imaging workflows, enabling seamless integration into clinical practice and research.
It provides a unified platform to query, interact with, and visualize AI-generated outputs, eliminating the need for separate AI-vendor workstations and reducing workflow fragmentation.
The plug-in supports diverse GenAI models including Large-Language Models (LLMs) for text generation, Vision-Language Models (VLMs) for text from combined text/image prompts, and AI-driven segmentation tools (like SAM) for ROI generation.
OpenAI's LLMs and VLMs, along with META's Segment Anything Model (SAM), were specifically integrated and tested to demonstrate capabilities.
OsiriXGPT aims to reduce reporting burdens, enhance diagnostic workflows, and improve communication of complex radiological information. It supports iterative "request-to-answer" interactions, allowing radiologists to refine AI outputs based on their expertise.
This integration facilitates evaluation of new GenAI models in research and clinical trials, particularly for complex tasks like WB-MRI reporting in cancer screening.
86.5%
Overall Average Classification Accuracy of VLMs
VLMs achieved 86.5% accuracy in identifying anatomical structures across all patients and structures, with some variability observed.
OsiriXGPT Workflow Overview
Radiologist uses OsiriXGPT
→
Input Text/Image/ROI Prompt
→
OsiriXgrpc API transfers data
→
Python Environment (GenAI Inference)
→
AI-Generated Output (Text/ROI)
→
Visualize in OsiriX Viewer
| Feature | OsiriXgrpc | Traditional AI Platforms |
|---|---|---|
| Integration | Seamlessly within OsiriX DICOM viewer | Often requires separate workstation/interface |
| Flexibility | Supports any gRPC-compatible language (e.g., Python) | Vendor-specific proprietary software |
| Cost/Access | Open-source, low-cost, adaptable for LMICs | High procurement costs, often individual purchases |
| Workflow | Unified 'scan-to-scan' reporting workflow | Fragmented, potential workflow disruption |
| Regulatory Pathway | Simplifies certification (only AI model needs approval) | AI model & platform require separate approvals |
WB-DWI SAM Segmentation for Prostate Cancer
The plug-in incorporated a fine-tuned SAM model for semi-automatic delineation of whole-body diffusion-weighted MRI (WB-DWI) lesions in oncology.
Key Findings:
- Good agreement in median ADC values observed with high ICC > 0.9.
- CoV of 4.05% for APC and 5.67% for myeloma, consistent with test-retest studies.
- 54/63 lesions demonstrated good DSC > 0.5.
Conclusion: This demonstrates the flexibility of OsiriXgrpc to integrate custom AI models, providing accurate and efficient segmentation for clinical research.
Quantify Your AI Impact
Estimate the potential time savings and cost efficiencies by integrating AI-powered segmentation tools into your radiology workflow.
Annual Savings
$0
Hours Reclaimed Annually
0
Your AI Implementation Roadmap
A strategic phased approach to integrating OsiriXgrpc and GenAI into your clinical and research workflows, ensuring successful adoption and maximum impact.
Phase 1: Initial Integration & Benchmarking
Successful implementation of OsiriXgrpc API with OpenAI LLMs, VLMs, and SAM for basic functionalities and performance testing.
Phase 2: Workflow Optimization & Usability Studies
Quantify reporting time reduction, interaction steps, and cognitive load. Conduct usability studies with radiologists using validated metrics.
Phase 3: Advanced AI Model Integration & Clinical Validation
Incorporate additional custom AI models, conduct multi-reader evaluations, and validate fully volumetric disease segmentation in diverse clinical scenarios.
Phase 4: Regulatory Approval & Real-world Adoption
Establish most effective display/interpretation methods, integrate model confidence scores, and pursue regulatory approval for clinical deployment.
Ready to Transform Your Radiology Workflow?
Leverage the power of OsiriXgrpc to integrate cutting-edge AI directly into your diagnostic imaging environment. Book a personalized consultation to explore how our solution can meet your specific needs and drive innovation.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat