NiranAI - Intelligent Diagnostic Workstation

Institution Management

• Maintain a lookup table with institution details
• Keycloak Group Management:
• Create Keycloak Groups for each institution using Keycloak APIs.
• Map Institution IDs from the database to the corresponding Keycloak Group IDs for backend integration.

Role-based access control (RBAC)

• Create roles as admin and radiologist
• Create resources/scopes/permissions in Keycloak:

• basic-viewer: Provides basic viewing permissions. (default features and tools of OHIF viewer along with manual segmentation)
• ai-segmentation: Grants access to AI-based segmentation features.
• ai-llm-diagnosis: Enables AI-based diagnostic tools.
• ai-dr
• ai-mammo
• ai-cervical

• Role Mapping:

• Assign roles to users through Keycloak Admin Console or Keycloak API.
• A user interface in application to create/modify users with roles. (future versions)

• Front-End Role Management

• Decode Authentication on the front-end to extract roles and institutional details from the token payload.
• Implement feature gating based on the decoded roles (enable/disable of AI based feature icons based on roles)

• Role-Based AI Service Execution on image file upload

• Automatically execute relevant AI Services (based on modality) when a DICOM file is uploaded to the server. (A configuration related settings can be set whether to auto execute AI services)

• Clinical Study Assignment for analysis

• Assign clinical studies to authorized users based on their roles and permissions. Images uploaded by a clinician should automatically be assigned to that clinician. For integration with external systems, have to implement a different logic to handle data mapping and assignment to a pool of clinicians.

File Handling & DICOM Conversion - Enable File Uploads in multiple formats:

• Integrate DICOM Conversion Services with capture of mandatory metadata (jpg, png, nifi). Customized the image upload feature of OHIF-Viewer.
• Include an option to Anonymize Datasets before processing.
• Integrate Image Anonymization Services

CT Image Analysis:

• Modality based page rendering to add an Icon in the Header to invoke Auto Full Slice Segmentation, Region-Based Prompt Segmentation and Slice-Specific Diagnosis
• Integrate Segmentation services and Slice-Specific Diagnosis services
• Resolve Issues: Fix dynamic segment creation in the Region-Based Prompt Segmentation service.
• Enhancement: Initiate Region-Based Prompt Segmentation asynchronously upon mouse release of the draw tool.
• Asynchronously call Segmentation Services in the background to maintain viewer interactivity. How the service in background updates in case region based prompt segmentation? A UI for scheduled services?
• Display Slice-Specific Diagnosis in the Right Panel, preserving previously recorded diagnoses for slices?

Fundus Image Analysis:

• Modality based page rendering to add an Icon in the Header to invoke the DR Discovery AI Service.
• Generate a DICOM SR (Structured Report) with the Diagnosis Results.

Mammogram Image Analysis:
Modality based page rendering to add an Icon in the Header to invoke the Disease Discovery AI Service.
Generate a DICOM SR Report with Diagnoses.

Cytology Image Analysis:
Modality based page rendering to add an Icon in the Header to invoke the Disease Discovery Service.
Generate a DICOM SR Report with Diagnosis Outcomes.

Deployment & Installation
• Hosting: Deploy the Application on a Public IP for external access.
• PWA Installation: Enable installation as a Progressive Web Application (PWA) for mobile access.

Two versions –
• NirnAI Enterprise Suite ….. for Hospital deployment hosting at hospital-site
• NirnAI Community Suite ….. for individual practitioners and public users hosting managed by C-DAC