Non-rigid Registration for Two-photon Imaging Using Triangulation and Piecewise Affine Transformation

Area of Science:

• Neuroscience
• Biomedical Imaging
• Computational Biology

Background:

• Multi-session two-photon (2p) imaging is crucial for studying neural mechanisms over time.
• Complex misalignments and data variance in 2p recordings hinder cross-session analysis.
• Existing registration tools struggle with multi-session, non-rigid deformations.

Purpose of the Study:

• To develop an accurate and efficient non-rigid registration method for multi-session 2p imaging in mice.
• To address the challenge of aligning neural images acquired across different experimental sessions.
• To facilitate discoveries in neuroscience by improving 2p data analysis.

Main Methods:

• Developed a novel non-rigid registration technique based on image triangulation and piecewise affine transformation (TPAT).
• Implemented TPAT to handle complex misalignments and large deformations in 2p imaging data.
• Evaluated the method's performance using benchmark tests for non-rigid neural image registration.

Main Results:

• The TPAT method demonstrated high accuracy and efficiency in non-rigid registration of 2p imaging data.
• Both automatic and semi-automatic operation modes of TPAT achieved excellent performance.
• Successfully addressed significant misalignment issues common in multi-session 2p recordings.

Conclusions:

• The developed TPAT method provides a robust solution for multi-session 2p image registration.
• This advancement is expected to accelerate discoveries in neural mechanisms by enabling reliable cross-session analysis.
• TPAT represents a significant step forward for researchers utilizing longitudinal 2p imaging data.