Enhancing volumetric microscopy with blind computational correction of spatially variant aberrations using neural...

Area of Science:

• Biomedical imaging
• Computational microscopy
• Optical engineering

Background:

• High-resolution imaging over large fields of view (FOV) is crucial for studying biological processes.
• Optical aberrations, especially in large-FOV systems, degrade image quality, limiting biological insights.
• Current aberration correction methods are often impractical, requiring hardware modifications or extensive calibration.

Purpose of the Study:

• To develop a computational method for correcting spatially varying optical aberrations in large-FOV microscopy.
• To enable high-resolution 3D reconstruction of biological structures without hardware changes or training data.
• To enhance the accessibility of advanced microscopy techniques for biological research.

Main Methods:

• A self-supervised algorithm was developed to simultaneously estimate aberrations and reconstruct 3D sample structure from single blurred images.
• The method does not require point spread function (PSF) calibration or large training datasets.
• Implementation was demonstrated on optically cleared mouse neurons/vasculature (OPM) and in vivo mouse retinal vasculature (AOSLO).

Main Results:

• The algorithm successfully predicted and corrected aberrations across multi-millimeter FOVs.
• Image contrast was enhanced in low signal-to-noise regions, revealing obscured structural details.
• High-resolution imaging of neuronal and vascular systems was achieved, demonstrating the method's effectiveness.

Conclusions:

• The self-supervised algorithm offers a practical solution for correcting field-dependent aberrations in large-scale biological imaging.
• This approach overcomes limitations of existing methods, making advanced microscopy more accessible.
• The technique facilitates detailed visualization of complex biological systems without compromising experimental workflows.