Custom 3D fMRI Registration Template Construction Method Based on Time-Series Fusion.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Standard medical image registration relies on MRI templates, lacking a 3D fMRI standard template.
• Registration to standard space introduces significant errors for subjects with non-standard brain structures.
• Current individual space registration methods for fMRI are subjective and dependent on researcher experience.

Purpose of the Study:

• To propose and validate a novel method for constructing a 3D custom functional MRI (fMRI) template.
• To enhance the accuracy and reliability of fMRI image registration by utilizing an individualized template.
• To address the limitations of existing standard templates and subjective individual space registration techniques.

Main Methods:

• Preprocessing of fMRI data.
• Registration and fusion of 2D slices from multiple time points to form image sequences for each brain layer.
• Feature point matching between adjacent time points and recursive mapping to a common space for template construction.

Main Results:

• A 3D customized fMRI template was successfully constructed with individual pertinence.
• Comparison with classic registration algorithms demonstrated improved registration accuracy using the custom fMRI template versus standard spaces.
• The proposed method leverages correlation information within sequence data for robust registration.

Conclusions:

• The developed method for constructing a 3D custom fMRI template significantly improves registration accuracy.
• Individualized fMRI templates offer a more reliable alternative to standard templates for diverse brain structures.
• This approach enhances the precision of medical image registration in neuroimaging studies.