Related Experiment Video
Updated: Oct 4, 2025
02:09
Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function
Published on: April 12, 2024
712
Long-range medical image registration through generalized mutual information (GMI): towards a fully automatic
Vinicius Pavanelli Vianna1, Luiz Otavio Murta1,2
1Department of Physics, University of Sao Paulo, Ribeirao Preto, Brazil.
Physics in Medicine and Biology
|February 7, 2022
Summary
Generalized Mutual Information (GMI) using Tsallis entropy enhances medical image registration by overcoming local minima issues common with standard Mutual Information (MI). This improves affine registration accuracy and robustness for brain MRI scans.
Area of Science:
- Medical Image Analysis
- Computational Neuroscience
- Information Theory
Background:
- Mutual Information (MI) is a standard metric for medical image registration.
- Standard MI can fail with large image transformations due to local minima.
- Affine registration requires robust similarity metrics to handle complex transformations.
Purpose of the Study:
- To propose and evaluate Generalized Mutual Information (GMI) using Tsallis entropy for improved affine medical image registration.
- To address the limitations of standard MI in handling large image transformations.
- To enhance the robustness and accuracy of brain MRI registration.
Main Methods:
- Assessed GMI metric output space with separable affine transforms (translations, rotations, scaling, skewness).
- Utilized 3D visualization of gradient and contour curves for data evaluation.
- Employed a simulated gradient descent algorithm and Monte Carlo simulations on brain MRI data (HCP dataset).
Main Results:
- GMI demonstrated significantly prolonged registration ranges without local minima.
- Achieved 100% registration capability for translations, scaling, and skewness; 88.2% for rotations.
- Tsallis entropy-based GMI achieved 99.75% success in Monte Carlo simulations, compared to 56.5% for Shannon entropy.
Conclusions:
- Tsallis entropy-based GMI significantly improves brain MRI affine registration.
- GMI overcomes local minima issues, enabling long-range translation registration.
- The improved gradient space facilitates lower-cost interpolation and faster registrations.
