Help & Documentation

Help Home
User Guide

Pre-Graded Image Uploads

This guide explains how to upload images that have already been graded, either by AI systems or by human graders. The pre-graded upload process consists of two main steps: uploading the images and importing the corresponding grades.

Overview

Pre-graded uploads allow you to import existing grading data into the system for review, analysis, or training purposes. This is particularly useful for: - Importing grades from external AI systems - Migrating data from legacy systems - Creating training datasets with known ground truth - Benchmarking AI performance against human graders - Quality assurance and review processes

Accessing Pre-Graded Uploads

Required Permissions

  • Admin: Full access to all pre-graded upload features
  • Data Manager: Can upload pre-graded images for assigned lab units
  • File Uploader: Can upload pre-graded images
  • Optometrist: Limited access based on permissions
  1. Upload Images: Navigate to Direct Uploads → Pre-Graded Upload
  2. Import Grades: Navigate to Direct Uploads → Pre-Graded Grades

Pre-Graded Image Upload Process

Step 1: Prepare Your Images

  1. Collect the images that have been pre-graded
  2. Ensure images are in supported formats (JPG, PNG)
  3. Organize images by hospital, lab unit, and disease type
  4. Check that images meet size requirements (typically max 5MB per file)

Step 2: Upload Pre-Graded Images

  1. Navigate to Direct Uploads → Pre-Graded Upload
  2. Fill in the required information:
  3. Hospital: Select the hospital where images originated
  4. Lab Unit: Select the specific lab unit
  5. Camera: Choose the camera type used
  6. Disease: Select the disease type (DR, Glaucoma, AMD, etc.)
  7. Area: Select the anatomical area
  8. Dataset Label: Optional label to identify this batch
  9. Mydriatic: Indicate if images were taken with mydriasis
  10. Select the image files to upload
  11. Click Upload to process the images

Step 3: Automatic Processing

When images are uploaded: 1. The system creates a job to track the upload process 2. Each image is stored with a unique identifier 3. Images are automatically marked as pre-graded 4. A verification record is created for each image 5. Grading tasks are created for each image

Pre-Graded Grade Import Process

Step 1: Prepare Grade Data

Create an Excel file with the following structure:

Required Columns

  • image_name: The filename of the graded image (must match uploaded files)
  • {role}_grade: Grade value for the specific role
  • For resident grades: resident_grade
  • For resident2 grades: resident2_grade
  • For AI grades: ai_grade

Optional Columns

  • {role}_remarks: Additional notes or comments
  • For resident grades: resident_remarks
  • For resident2 grades: resident2_remarks
  • For AI grades: ai_remarks
  • ai_probability: Confidence score for AI grades (only for AI grades)

Step 2: Import Grades

  1. Navigate to Direct Uploads → Pre-Graded Grades
  2. Select the grader type:
  3. Resident: For resident doctor grades
  4. Resident2: For resident2/ophthalmologist grades
  5. AI: For artificial intelligence grades
  6. Fill in the required information:
  7. Hospital: Select the hospital (must match image upload)
  8. Lab Unit: Select the lab unit (must match image upload)
  9. Disease: Select the disease type (must match image upload)
  10. Grader: Select the user who performed the grading (for human grades)
  11. AI Model: Select or create AI model information (for AI grades)
  12. Upload the Excel file containing grade data
  13. Click Upload to process the grades

Step 3: Grade Mapping

If the system encounters grade values it doesn't recognize: 1. A mapping interface will appear 2. You'll see unrecognized grade values from your file 3. Map each value to the appropriate system grade,=. This is to be done by the user 4. The system may auto-map some values based on text matching 5. Complete the mapping and click Import

Step 4: Processing Results

The system will: 1. Validate all grade mappings 2. Match grades to uploaded images by filename 3. Create grade records for each image 4. Update task states for human grades 5. Create consensus grades for resident2 grades 6. Display a summary of successful and failed imports

Supported Grade Types

Resident Grades

  • Who: Medical residents in training
  • Requirements: User must have "resident" or "ophthalmologist" role
  • Process: Grades are imported and task states are updated
  • Use Case: Training evaluation, quality assessment

Resident2 Grades

  • Who: Senior ophthalmologists and resident2 members
  • Requirements: User must have "ophthalmologist" role
  • Process: Grades are imported and consensus is calculated
  • Use Case: Gold standard grades, reference data

AI Grades

  • What: Grades generated by artificial intelligence systems
  • Model Information: Can select existing AI models
  • Process: Grades are imported with model metadata
  • Use Case: AI validation, model comparison, research

Grade Matching Process

The system follows these steps to match grades to images:

  1. Filename Matching: Looks for exact filename matches (case-insensitive)
  2. Location Filtering: Filters by hospital, lab unit, and disease
  3. Pre-Graded Filter: Only matches images marked as pre-graded
  4. Most Recent: If multiple matches exist, uses the most recent upload
  5. Task Creation: Ensures grading tasks exist for matched images

Duplicate Handling

Image Duplicates

  • Pre-graded images use the same duplicate detection as regular uploads
  • MD5 hash is calculated and checked against existing images
  • Duplicate images are stored in a separate folder
  • Upload status shows "Duplicate file" for duplicates

Grade Duplicates

  • New grades can override existing grades for the same grader type
  • Multiple different grader types can be applied to the same image
  • AI grades can be imported multiple times with different models
  • The system tracks all grade changes with timestamps

Quality Assurance

Validation Checks

  • Filename Validation: Ensures all grade entries have matching images
  • Grade Validation: Verifies grade values are within allowed ranges
  • Permission Validation: Checks user permissions for lab units
  • Format Validation: Ensures Excel file has correct structure

Error Handling

  • Missing Images: Grades without matching images are flagged as errors
  • Invalid Grades: Grade values that can't be mapped are rejected
  • Permission Issues: Attempts to upload to unauthorized lab units are blocked
  • Format Errors: Invalid Excel formats are rejected with helpful messages

Best Practices

File Preparation

  1. Consistent Naming: Use consistent filenames for images and grade files
  2. Complete Data: Ensure all required fields are filled in
  3. Valid Grades: Use grade values that exist in the system
  4. Test Samples: Test with a small sample before large imports

Upload Process

  1. Upload Images First: Always upload images before importing grades
  2. Verify Matching: Check that image names match exactly
  3. Review Results: Review import results for any errors
  4. Handle Errors: Address any mapping or matching issues promptly

Data Management

  1. Organize by Batch: Use dataset labels to organize uploads
  2. Document Sources: Keep track of where grades originated
  3. Version Control: Use AI model versioning for AI grades
  4. Quality Review: Periodically review imported grades

Troubleshooting Common Issues

Image Not Found

  • Problem: Grade references an image that wasn't uploaded
  • Solution: Ensure all images are uploaded before importing grades
  • Check: Verify filenames match exactly (including extensions)

Grade Mapping Errors

  • Problem: System doesn't recognize grade values
  • Solution: Use the mapping interface to match values to system grades
  • Prevention: Use standard grade terminology in your files

Permission Denied

  • Problem: User can't upload to selected lab unit
  • Solution: Contact administrator for appropriate permissions
  • Check: Verify you're selecting the correct lab unit

Import Failures

  • Problem: Partial or complete import failure
  • Solution: Review job details for specific error messages
  • Common Causes: Invalid formats, missing data, permission issues

Integration with Grading Workflows

Task Creation

  • Pre-graded uploads automatically create grading tasks
  • Tasks are marked with appropriate grader assignments
  • Human grades update task states automatically
  • Resident2 grades trigger consensus calculations

Review Process

  • Pre-graded images appear in regular grading queues
  • Graders can see pre-assigned grades when evaluating
  • Discrepancies between graders are tracked
  • Quality metrics include pre-graded data

Security and Privacy

  • Access Control: Only authorized users can upload pre-graded content
  • Audit Trail: All uploads and imports are logged
  • Data Integrity: Verification ensures data consistency
  • Privacy Compliance: Patient data is handled according to regulations

Getting Help

For pre-graded upload issues: 1. Check the error messages in the job details 2. Verify your data format matches requirements 3. Ensure you have appropriate permissions 4. Contact your system administrator for assistance