X-Ray2EM: Uncertainty-Aware Cross-Modality Image Reconstruction from X-Ray to Electron Microscopy in Connectomics.

Area of Science:

• Neuroscience
• Biophysics
• Computational Biology

Background:

• Connectomics, the study of neural connections, traditionally relies on volume electron microscopy (EM), a complex and destructive process.
• EM involves thin sectioning, imaging, alignment, and reconstruction, which are time-consuming and technically challenging.
• Hard X-ray imaging offers a non-destructive alternative compatible with thick tissues, promising faster acquisition and intrinsic alignment.

Approach:

• We propose an uncertainty-aware 3D reconstruction model to translate X-ray images into EM-like images.
• This model enhances the segmentation quality of neural membranes, a critical feature for connectomics.
• The approach aims to bridge the resolution gap between X-ray microscopy and EM.

Key Points:

• X-ray imaging eliminates the need for fragile thin sectioning required in EM.
• The developed model significantly improves membrane segmentation in X-ray microscopy data.
• This work demonstrates the potential of X-ray based methods for high-resolution connectomics.

Conclusions:

• Our uncertainty-aware reconstruction model facilitates higher-quality connectomics from X-ray data.
• This approach paves the way for simpler, faster, and more accurate X-ray based connectomics pipelines.
• It offers a promising alternative to EM for comprehensive brain imaging and neuronal function studies.