NucleiSky enables cross-scale multimodal registration of microscopy data using nuclei constellations

Area of Science:

• Biophysics
• Computational Biology
• Microscopy

Background:

• Integrating multi-modal and multi-scale microscopy data is crucial for understanding biological systems.
• Accurate image registration across different imaging modalities, settings, and scales remains a significant challenge.

Purpose of the Study:

• To introduce NucleiSky, a novel framework for robust microscopy image registration.
• To enable precise alignment of images acquired with different microscopy techniques and magnifications.

Main Methods:

• NucleiSky utilizes the spatial arrangement of nuclei as an intrinsic biological fingerprint.
• Images are represented as constellations of centroids and aligned using geometric algorithms and spatial consensus scoring.
• The framework supports 2D and 3D localization, and can utilize non-nuclear landmarks.

Main Results:

• NucleiSky accurately localizes query regions within larger reference images using minimal landmarks (as few as five nuclei).
• Demonstrated capability to map high-magnification fields of view within low-magnification scans and across additional channels.
• Successfully applied for live brightfield-to-fixed registration and guided microscope re-targeting.

Conclusions:

• Local landmark geometry serves as a robust intrinsic spatial fingerprint for image registration.
• NucleiSky facilitates accurate localization and registration across diverse imaging scales, modalities, and microscopy platforms.
• The open-source NucleiSky package offers a versatile solution for microscopy image analysis.