Skip to main content

Volume Similarity Statistics

The Volume Similarity Statistics tool calculates similarity metrics between two volume datasets. This is useful for comparing registered images, evaluating alignment quality, or assessing changes between time points.

Accessing the Tool

  1. Navigate to the Measure tab in the ribbon
  2. Click the Volume Similarity Statistics button in the Statistics section

Overview

Calculate similarity statistics between two volume datasets.

Prerequisites

Two volume datasets must be loaded in the project. For meaningful results, volumes should:

  • Have the same dimensions and spacing, or be properly registered
  • Represent the same anatomical region or structure
  • Use compatible intensity scales

User Interface

Description

Displays: "Calculate similarity statistics between two volume datasets."

Volume Objects

ControlDescription
Input 1Select the first volume dataset for comparison
Input 2Select the second volume dataset for comparison

Both dropdowns list all volume datasets in the project.

Similarity Statistics Table

Displays calculated similarity metrics in a name-value table:

MetricDescription
Cross correlation (CC)Measures linear relationship between intensities
Local normalized cross correlation (LNCC)Locally normalized version of cross correlation
Correlation ratio (XY)Functional dependence of Y on X
Correlation ratio (YX)Functional dependence of X on Y
Mutual information (MI)Information-theoretic similarity measure
Normalized mutual information (NMI)Normalized version of mutual information
Kappa statistic (K)Agreement measure accounting for chance
Cosine of normalized gradient fieldStructural similarity based on gradients
Joint entropy (JE)Combined entropy of both images
Sum of sq. intensity differences (SSID)Total squared intensity differences
Mean sq. error (MSE)Average squared difference per voxel
Sum of absolute differences (SAD)Total absolute intensity differences
Sum of sq. differences (SSD)Sum of squared differences
CovarianceJoint variability measure
Peak signal-to-noise ratio (PSNR)Ratio of signal power to noise power
Label consistency (LC)Agreement between discrete labels

Toolbar

ButtonDescription
Export...Export statistics to a file
UpdateRecalculate statistics

Similarity Metrics

Correlation-Based Metrics

MetricRangeInterpretation
Cross Correlation (CC)-1 to 11 = perfect positive correlation
LNCC-1 to 1Robust to intensity variations
Correlation Ratio0 to 11 = perfect functional dependence

Information-Theoretic Metrics

MetricRangeInterpretation
Mutual Information (MI)≥ 0Higher = more shared information
Normalized MI (NMI)0 to 2Normalized for comparison
Joint Entropy (JE)≥ 0Lower = more similar

Difference-Based Metrics

MetricRangeInterpretation
MSE≥ 00 = identical images
SAD≥ 00 = identical images
SSD≥ 00 = identical images
PSNRdB scaleHigher = more similar
Choosing the Right Metric
  • Cross Correlation (CC): Best for images with linear intensity relationships
  • Mutual Information (MI): Robust for multi-modal image comparison (e.g., CT vs MRI)
  • MSE/PSNR: Simple metrics suitable for images with similar intensity ranges
  • Normalized metrics: Preferred when comparing images with different intensity scales

Workflow

Basic Comparison

  1. Ensure at least two volume datasets are loaded
  2. Open the Volume Similarity Statistics tool
  3. Select Input 1 from the first dropdown
  4. Select Input 2 from the second dropdown
  5. Click Update to calculate statistics
  6. Review the similarity metrics in the table

Registration Quality Assessment

  1. Perform image registration between two volumes
  2. Open the Volume Similarity Statistics tool
  3. Compare the registered volume with the reference
  4. Higher correlation and MI values indicate better alignment

Change Detection

  1. Load baseline and follow-up volumes
  2. Calculate similarity statistics
  3. Lower similarity values may indicate changes between acquisitions

Export

Click Export... to save the similarity statistics to a file for documentation or further analysis.


Use Cases

Registration Validation

Verify that image registration achieved good alignment by checking correlation and mutual information values.

Quality Control

Compare acquired images against reference standards to ensure imaging consistency.

Longitudinal Studies

Quantify changes between time points by tracking similarity metrics over time.

Algorithm Comparison

Evaluate different processing algorithms by comparing their outputs to a reference.


Technical Notes

  • Both volumes should have the same dimensions for accurate comparison
  • Volumes are compared voxel-by-voxel based on spatial correspondence
  • For unregistered volumes, consider performing registration first
  • Some metrics (CC, LNCC) are more robust to intensity scaling differences