Registration

Volvicon provides two registration methods for aligning multiple datasets: Landmark Registration for manual point-based alignment and Global Registration for automatic alignment using surface or image features.

---

Landmark Registration

Landmark Registration aligns datasets using corresponding points placed manually on the fixed (reference) and moving (to be aligned) objects.

Accessing the Tool

1. Navigate to the Image tab in the ribbon.
2. Click Landmark Registration in the Transform section.

Concepts

Term	Description
Fixed object	The reference dataset that remains stationary
Moving object(s)	Datasets that will be transformed to align with the fixed object
Landmark	A point placed on an identifiable anatomical or geometric feature

Requirements

• Minimum 3 landmark pairs are required for registration.
• Landmarks must be placed on corresponding features in both datasets.
• More landmarks generally improve registration accuracy.

Workflow

1. Open the Landmark Registration tool.
2. Select the fixed object (reference) and moving object(s).
3. Place landmarks on the fixed object by clicking identifiable features.
4. Place corresponding landmarks on the moving object in the same order.
5. Verify that landmark pairs correspond correctly (same anatomical/geometric feature).
6. Optionally enable Global registration fine-tuning for automatic refinement.
7. Click Apply to perform the registration.

Landmark Table

The tool displays a table with landmark coordinates:

Column	Description
#	Landmark index
X, Y, Z (mm)	World coordinates of the landmark point

Use the table toolbar to:

• Add new landmarks
• Remove selected landmarks
• Clear all landmarks
• Import/Export landmarks from/to files
• Move Up/Down to reorder landmarks

Registration with Fine-Tuning

Enable Perform global registration after landmark alignment to:

1. Apply initial alignment from landmark correspondence.
2. Automatically refine the alignment using surface or image features.
3. Achieve higher accuracy than landmark-only registration.

Accuracy Assessment

After registration, the tool displays:

Metric	Description
Landmark RMSE	Root Mean Square Error of landmark positions after registration
Registration RMSE	Overall alignment error (if global refinement is enabled)

---

Global Registration

Global Registration automatically aligns datasets without manual landmark placement, using surface geometry or image intensity features.

Accessing the Tool

1. Navigate to the Image tab in the ribbon.
2. Click Global Registration in the Transform section.

Registration Types

The appropriate registration method is selected automatically based on object types:

Fixed Object	Moving Object	Method Used
Surface/Mask/Mesh	Surface/Mask/Mesh	Surface-based registration
Volume	Volume	Image-based registration

Surface-Based Registration

Aligns objects using surface geometry. Suitable for:

• Surface meshes (STL, CAD models)
• Mask objects (uses extracted surface)
• Volume meshes

Algorithms

Algorithm	Description
Point matching	Iterative closest point (ICP) algorithm that minimizes point-to-point distances
Feature matching	Aligns based on geometric feature descriptors
Point + Feature matching	Uses feature matching for initial alignment, then point matching for refinement

Point Matching Parameters

Parameter	Description
Mode	Transformation type: Rigid, Similarity, or Affine
Maximum landmarks	Maximum number of landmarks sampled from the surface
Match centroids	If enabled, align object centroids before registration

Feature Matching Parameters

Parameter	Description
Global iterations	Number of times the algorithm runs
Overlap radius	Optimization radius around each point
Curvature threshold	Threshold for feature detection (smaller = higher accuracy)

Point + Feature Matching Parameters

Combines both parameter sets for a two-stage registration process.

Image-Based Registration

Aligns volumes using voxel intensity information. Suitable for:

• CT-to-CT alignment
• MRI-to-MRI alignment
• Multi-modal registration (with appropriate configuration)

Registration Methods

Method	Description
Rigid (6 DOF)	6 degrees of freedom: 3 rotations + 3 translations
Affine (12 DOF)	12 degrees of freedom: includes rotation, translation, scaling, and shearing

Parameters

Parameter	Description
Number of iterations	Maximum iterations for the optimization algorithm
Learning rate	Step size for the optimization process
Relaxation factor	Controls convergence speed and stability

Workflow

1. Open the Global Registration tool.
2. Select the fixed object (reference).
3. Select moving object(s) to be aligned.
4. Choose registration parameters (or use defaults).
5. Optionally enable Calculate RMSE to assess registration quality.
6. Click Apply to perform the registration.

Output

After successful registration:

• Moving objects are transformed to align with the fixed object.
• The transformation matrix is applied to all selected moving objects.
• RMSE values are displayed if enabled.

---

Best Practices

Landmark Placement

• Choose features visible in both datasets.
• Use anatomical landmarks (e.g., bone prominences, vessel bifurcations).
• Distribute landmarks across the region of interest.
• Avoid placing landmarks on noise or artifacts.

Pre-Processing

• Coarse manual alignment: Position datasets approximately before automatic registration.
• Cropping: Remove irrelevant regions to improve registration speed and accuracy.
• Smoothing: Reduce noise for better feature matching.

Validation

• Visually inspect alignment in orthogonal views.
• Use overlay or checkerboard visualization to compare aligned datasets.
• Check RMSE values; lower values indicate better alignment.

Medical Disclaimer

Registration results should be verified by qualified professionals before use in clinical decision-making. Automatic registration algorithms may produce suboptimal results in certain cases.

---

Scripting

Registration operations are available via the Python scripting API using the RegistrationOperations class.

See the RegistrationOperations API Reference for available methods including global_registration() and landmark_registration().