MULTIMODAL MICROSCOPY IMAGE ALIGNMENT USING SPATIAL AND SHAPE INFORMATION AND A BRANCH-AND-BOUND ALGORITHM.

Area of Science:

• Neuroscience
• Microscopy
• Computational Biology

Background:

• Multimodal microscopy is vital for systems and molecular neuroscience.
• Aligning images from different modalities (e.g., gene expression, calcium signals) is challenging, especially with limited cell overlap.
• Existing image registration methods struggle with sparse cell populations common in multimodal experiments.

Purpose of the Study:

• To develop an efficient and globally optimal algorithm for aligning multimodal microscopy images.
• To address the challenge of cell subset matching in image registration.
• To enhance the integration of complementary cellular information from different imaging modalities.

Main Methods:

• Formulating multimodal microscopy alignment as a cell subset matching problem.
• Implementing an efficient and globally optimal branch-and-bound algorithm for point cloud subset alignment.
• Utilizing cell shape and location data to compute matching likelihood and prune the search tree.
• Seeding image deformation fields with aligned cell subsets for final registration.

Main Results:

• The proposed branch-and-bound algorithm achieves globally optimal solutions for rotational alignment of cell subsets.
• The method effectively utilizes complementary cell information to improve matching accuracy.
• The framework demonstrates superior matching quality compared to state-of-the-art histology alignment methods.
• The algorithm is faster than manual alignment, significantly increasing experimental throughput.

Conclusions:

• This novel approach provides a robust solution for multimodal microscopy image alignment.
• The cell subset matching strategy overcomes limitations of traditional registration methods.
• The developed algorithm enhances the efficiency and accuracy of integrating diverse cellular data, advancing neuroscience research.