Semi-automatic stitching of filamentous structures in image stacks from serial-section electron tomography

Area of Science:

• * Cell Biology
• * Computer Science
• * Biophysics

Background:

• * Electron tomography enables 3D reconstruction of filamentous structures like microtubules (MTs) in cell division.
• * Large-scale reconstructions of spindles require stitching and aligning serial tomograms, which is challenging due to data imperfections.
• * Previous fully automatic approaches struggle with physical distortions and low signal-to-noise ratios in large datasets.

Purpose of the Study:

• * To develop an interactive, semi-automatic software solution for large-scale stitching and alignment of filamentous structures in serial-section electron microscopy data.
• * To enable accurate 3D modeling of microtubules (MTs) in mitotic and meiotic spindles despite data imperfections.
• * To provide a user-friendly workflow balancing quality and manual effort for analyzing complex biological samples.

Main Methods:

• * A software-assisted workflow combining automatic matching, visual validation, and interactive correction of landmarks and filament matches.
• * Development of specialized visualization and interaction tools for real-time response and efficient user guidance through large datasets.
• * Iterative refinement process with close collaboration between computer scientists and cell biologists.

Main Results:

• * Successful large-scale 3D reconstruction of spindles where fully automatic workflows failed, requiring manual corrections.
• * Demonstration of a specialized tool capable of processing large-volume, imperfect electron tomography data for filamentous structures.
• * Validation of the semi-automatic approach for tracing and connecting microtubules across tomogram boundaries.

Conclusions:

• * The developed software provides a unique and effective solution for the challenging task of large-scale stitching of filamentous structures in serial-section image stacks.
• * The interactive, semi-automatic approach is crucial for achieving high-quality 3D reconstructions from imperfect electron tomography data.
• * This methodology is applicable to various biological samples with 3D distributions of microtubules in different cellular contexts.