Gene Expression Data to Mouse Atlas Registration Using a Nonlinear Elasticity Smoother and Landmark Points...

Area of Science:

• Computational biology
• Neuroscience
• Medical imaging

Background:

• Accurate registration of gene expression data to anatomical atlases is crucial for understanding spatial gene function.
• Existing image registration methods using linear regularization struggle with complex deformations and feature matching.

Purpose of the Study:

• To develop and evaluate a numerical algorithm for image registration using nonlinear elasticity regularization.
• To improve the accuracy of registering gene expression data to a 2D neuroanatomical mouse atlas.

Main Methods:

• A nonlinear elasticity regularization model was applied to allow for larger, smoother deformations.
• Optimality constraints on landmark point distances were enforced for enhanced feature matching.
• A matrix variable was introduced to approximate the Jacobian matrix for solving simplified Euler-Lagrange equations.

Main Results:

• The nonlinear elasticity model required fewer numerical corrections (e.g., regridding) compared to linear regularization.
• Improved ground truth rendering and increased mutual information were observed.
• Smaller landmark point distances and L2 dissimilarity measures were achieved compared to biharmonic regularization.

Conclusions:

• The proposed nonlinear elasticity regularization algorithm offers superior performance for gene expression data registration.
• This method enhances accuracy, reduces computational corrections, and improves feature matching in neuroanatomical atlas registration.