Annotation Project Tutorial (2D)

This tutorial will guide you through the steps to create a 2D annotation project. Follow the instructions carefully to set up your project successfully.

Step 1: Dataset Selection

1. Search

• Use the search bar to locate your dataset by name.
  • Type "CheXpert" in the search bar to find the CheXpert: Chest X-rays Dataset dataset.

2. Apply Filters

• Refine your search using filters:
  • Select filters such as:
    • Modality: Choose the data type (e.g., XRay, CT).
    • Anatomy: Filter by regions like Chest or Neuron.
    • Image Format: Select JPG, PNG, or DICOM.
    • Category: Choose the source (e.g., NIH).
    • Source Name: Specify the source name (e.g., "The Cancer Image Archive").

3. Select a Dataset

• Click the radio button for your chosen dataset.
  • Example: Select CheXpert: Chest X-rays Dataset.

5. Proceed

• Click Continue to move to the next step: Project Details.

Step 2: Project Details

1. Name Your Project

• Enter a meaningful name for your project.
  • Type Chest X-Ray Annotation Project.

2. Write a Description

• Provide an overview of the project's objective.
  • Write a description like:
    This project involves annotating chest X-rays from the CheXpert dataset to identify and label anomalies or patterns indicative of diseases.

3. Provide Annotator Instructions

• Write step-by-step instructions for annotators.
  • Include points such as:
    • Use the segmentation tools to annotate abnormalities.
    • Ensure all regions of interest are accurately labeled.
    • Refer to the annotation guidelines provided in the training module.

4. Add Tags

• Use tags to categorize the project.
  • Add tags like Chest X-Ray, CheXpert, Medical Imaging.

5. Configure Viewer Settings

• Keep the viewer set to 2D.

  • Ensure the 2D Viewer option is selected.

  

6. Enable Review Requirements

• Keep this enabled to ensure all annotations are reviewed.

7. Optional: Rate Assessments

• Enable this if annotators need to rate task difficulty.

8. Enable AI Assistance

• Turn this on if you want AI to assist in annotations.

9. Proceed

• Click Continue to move to the next step: Team.

Step 3: Team

1. Add Annotators

• Add team members via email.
  • Select a reader from the dropdown.
  • You can see owner of the project at email column.

2. Assign Roles

• Assign roles to team members:
  • Enable the features you want for the readers you want:
    • Annotator: For labeling data.
    • Reviewer: For reviewing annotations.
    • Manager: For managing the project.
    • Guest: For limited access.
    • Data Quality Control: For overseeing annotation quality.

3. Set Assignments

• Assign datasets to team members.
  • Click the assignment icon in the Annotator column to open the Assignment Modal.

4. Proceed

• Click Continue to move to Step 4: Labeling Configuration.

Assignment Modal

1. Select Images

• Assign images for annotation:
  • Select individual images or use Select All to assign all visible images.

2. Finalize Assignments

• Click Submit to save the assignments.
  • Use Reset to clear all selections.

Step 4: Labeling Configuration

1. Add Labeling Configuration

• Set up labels for the project.
  • Click + Labeling Configuration.

2. Optional: Add Measurement Form

• Define metrics for annotations.
  • Click + Measurement Form.

3. Optional: Add Image Evaluation Form

• Add criteria for evaluating image quality.
  • Click + Image Evaluation Form.

4. Finalize Configuration

• Click Create to save and finalize your project setup.

New Labeling Configuration Modal

1. Name

• Enter a descriptive name for the labeling configuration.
  • Type CheXpert: Chest X-Ray Anomaly Detection.

2. Problem Type

• Select the type of annotation problem:
  • Classification
  • Segmentation
  • Semantic Segmentation
  • Instance Segmentation
  • Object Detection

3. Annotation Type (Segmentation is selected at example)

• If you select Segmentation as the problem type, choose an annotation type:
  • Polygon: Use this for freehand or custom-shaped annotations.
  • Mask: Use this for pixel-level annotations to label areas with high precision.

4. Default Tool (If Mask is Selected)

• If Mask is chosen as the annotation type, select a default tool:
  • Brush: For painting regions manually with a brush-like tool.
  • Freehand Scissors: For drawing freehand outlines to define regions.
  • Correction Scissors: For modifying or correcting existing masks.

5. Allow Multiple Instances

• Enable this if multiple annotations of the same type are needed.
  • Example: Enable if annotating multiple lesions.

4. Add Classes

• Add categories for annotations:
  • Click + New Class and add background, lung.

5. Save

• Click Create to save the configuration.

New Measurement Form Modal

1. Name

• Enter a name for the measurement form that reflects its purpose.
  Example: CheXpert X-Ray Measurements

2. Preferred Series

• Specify the preferred dataset series if applicable.
  Example: CheXpert Chest X-Rays

3. Object List

• Add objects for which measurements will be made.
  Example: Lesions, Tumors

4. Measurement Set

• Click + Add Measurement Set to define the types of measurements required:
  • Name: Enter the name of the measurement.
    • Example: Tumor Length
  • Label: Provide a short label or abbreviation for the measurement.
    • Example: Len
  • Type: Select the measurement type from the dropdown options:
    • Length: For linear measurements.
    • Polygon: For freehand shape measurements.
    • Angle: For angular measurements.
    • Cobb Angle: Specific for spine curvature.
    • Mask: For segmented regions.
    • Bidirectional: For width and height measurements.
    • Rectangle: For bounding box measurements.
    • Point: For single coordinate annotations.

5. Create

• Once the details are configured, click Create to save the measurement form.

New Image Evaluation Modal

1. Table Name

• Enter a name for the evaluation table to describe its purpose.
  Example: CheXpert X-Ray Quality Evaluation

2. Group Name

• Add a group name to organize related evaluation questions.
  Example: Image Clarity

3. Add Questions

• Click + Add New Question to define criteria for evaluation.
  Examples of questions:
  • "Is the image properly centered?"
  • "Are the anatomical structures clearly visible?"
  • "Is there any visible noise or artifact?"

4. Add New Groups (Optional)

• Click + Add New Group to categorize questions further.
  Example:
  • Group Name: Anatomy Visibility
    Questions:
    • "Are the lungs fully visible?"
    • "Is the heart region clearly outlined?"

5. Save

• After adding all necessary groups and questions, click Save to finalize the image evaluation form.

Accessing Your Project

1. Locate Your Project

• The Annotation Projects dashboard displays all existing projects.
• Projects are shown as cards containing:
  • Project Name: Describes the project's purpose (e.g., Chest X-Ray Annotation Project).
  • Created Date: When the project was created.
  • Last Activity: The most recent action performed.

2. Identify the Project Card

• Locate the card corresponding to your desired project.
  Example: Look for "Chest X-Ray Annotation Project."

3. Click on the Project Card

• Click the project card to open it and view detailed options, such as annotations, settings, or project statistics.

4. Open the Project

• Click Open Project and select your preferred annotation mode:
  • Open with Annotation for 2D projects. (Click on this for example project)
  • Open with 3D Annotation for 3D-enabled datasets.

Annotation Viewer Page (2D)

1. User Mode Options

• Select the appropriate User Mode from the dropdown menu:
  • Annotator: Choose this if you will be labeling or annotating data. (Select Annotator option to make segmentations.)
  • Reviewer: Select this if your task is to review and validate annotations.
  • Manager: This role is for managing the project and its configurations.
  • Guest: Use this mode for limited access, such as observing project progress.

2. Select Labels for Annotation

• From the Label Selection panel, choose the appropriate label for the region you want to annotate:
  • Example Labels:
    • Normal Tissue: Use for areas with no abnormalities.
    • Pneumonia: Label suspected pneumonia regions.
    • Tumor: Mark areas suspected to contain tumors.

3. Use Segmentation Tools

• Access the segmentation tools from the toolbar:
  • Brush Tool: Use this tool for brush, freehand, scissor, threshold brush options. Adjust size, opacity, and brush threshold using the Brush Configs in the Segment Tools panel.
  • Eraser Tool: Correct any mistakes by erasing unwanted annotations.
• For Brush Tool Settings:
  • Adjust Opacity to control annotation visibility.
  • Use the Brush Radius slider to change the size of the brush.
  • Toggle Smart Segmentation for AI-assisted segmentation.

4. Draw the Labels

• With the selected tool and label:
  • Click and drag on the image to annotate the chosen region.
  • Ensure the labeled area corresponds to the selected label (e.g., Normal Tissue, Pneumonia, Tumor).

5. Save and Review

• Once the annotation is complete:
  • Click Save to finalize your work.
  • Use the Study Completed toggle to mark the current image as fully annotated.

6. Image Selection and Navigation

• Selections Panel:
  • Toggle the Study Completed switch to mark the current image as fully annotated once your work is finished.
  • Monitor progress through the Completed counter, which shows the number of completed images and the overall percentage.

• Image List:
  • Use the search bar to locate a specific image by its name or identifier.
  • Scroll through the list to select the image you want to annotate.
  • Click on an image to load it into the annotation viewer for labeling.

7. Saving Progress

• Always click Save after completing annotations on an image to ensure your work is stored.
• Use the Study Completed toggle to update the project's progress and indicate completed studies.

Toolbar Functions Overview

Explanation of each tool from left to right:

1. Reset

• Resets the image to its default zoom level and orientation.

2. Segment Tools

• Opens the segmentation tools panel for annotating and labeling regions in the image.

3. Hide Segments

• Toggles the visibility of existing annotations or segments on the image.

4. Stack Scroll

• Allows navigation through a series of stacked images or slices, useful for multi-frame datasets.

5. Zoom

• Adjusts the zoom level for closer inspection or a broader view of the image.

6. Pan

• Enables dragging of the image across the workspace for better positioning.

7. Erase Tool

• Activates the eraser to remove unwanted annotations or markings.

8. Fit to Window

• Scales the image to fit the available workspace entirely.

9. Invert

• Inverts the image colors, useful for enhanced visibility of certain features.

10. HFlip

• Flips the image horizontally.

11. VFlip

• Flips the image vertically.

12. Rotate Right

• Rotates the image 90 degrees clockwise.

13. Rotate Left

• Rotates the image 90 degrees counterclockwise.

14. Download

• Downloads the annotated image or the original image to your local device.

15. Reference Tool

• Opens the reference panel to compare the current image with a reference image.

16. Drag Probe

• Activates the drag probe for inspecting pixel values or other metadata.

17. Grid View

• Toggles the display of a grid overlay for alignment and precision during annotation.