Getting Started
- Obtain an access token using the authentication flow (see Authentication)
- Upload base64-encoded lab report documents using the POST endpoint
- Retrieve results either synchronously or via webhook notifications
Key Features
AI-Powered Analysis
- Advanced OCR for PDF and image document processing
- LOINC standard terminology for precise test result classification
- Machine learning models for accurate data extraction and validation
Flexible Processing
- Asynchronous processing (recommended) — Returns immediately with background processing
- Synchronous processing — Waits for complete analysis before returning results
- Webhook notifications for real-time status updates
Comprehensive Data
- LOINC code mapping for standardized test identification
- Structured sections with organized test results
- HIPAA compliant automatic de-identification
- Quality validation with human review flags for uncertain results
Technical Requirements
Document Specifications
- Maximum size: 10MB when base64-encoded
- Supported formats: PDF (preferred), JPEG, PNG
- Input format: base64-encoded string
- Quality: High resolution (300+ DPI) recommended for best OCR results
Processing Workflow
- Document Upload & Validation — System validates format compatibility
- OCR Processing — Extract text from PDF or image documents
- AI Analysis — Parse sections and identify individual test results
- LOINC Mapping — Map test names to standard LOINC codes when possible
- Validation — Flag uncertain results for human review
- Anonymization — Remove sensitive patient information per HIPAA guidelines
- Storage — Results stored with configurable retention policy
Analysis Report Data
Main Data
- Collection and result dates
- Original and standardized section names
- Test-specific notes and flags
- Processing status and error details
Test Results Structure
- Original test names from the lab report
- LOINC codes and common names when available
- Numeric values with units or text values
- Normal range indicators
- Human review requirements
LOINC Integration
- Sections: Mapped to panel-level LOINC codes (e.g., “58410-2” for CBC panel)
- Individual Tests: Specific LOINC codes for each test (e.g., “6690-2” for WBC count)
- Quality Assurance: Only confident mappings are included; uncertain mappings flagged for review
Document Guidelines
For optimal analysis results, ensure lab reports:- High quality — Use clear, high-resolution scans (300+ DPI)
- Complete pages — Include all pages of multi-page reports
- Proper orientation — Ensure documents are right-side up
- Good contrast — Text should be clearly readable against background
- Standard formats — PDF preferred for text-based reports, high-quality images for scanned documents
- Remove artifacts — Avoid shadows, folds, or obstructions that might interfere with OCR
Implementation
For detailed implementation examples, request/response schemas, and code samples, see:- Implementation Guide — Complete code examples and integration patterns
- API Reference — Detailed endpoint specifications and data types
- Asynchronous Processing — Real-time notifications and webhook implementation
LOINC Code Support
The API supports mapping to standard LOINC terminology for both sections and individual test results. By default, mappings are only included when the confidence level is high to ensure accuracy. Uncertain mappings are flagged withrequire_human_review: true for manual verification.
LOINC codes follow the standard 6-digit format (e.g., “6690-2”, “58410-2”) with corresponding common names.