CNN-based approach for 3D artifact correction of intensity diffraction tomography images.

Area of Science:

• Biomedical Imaging
• Developmental Biology
• Image Processing

Background:

• Limited-angle tomographic imaging frequently causes elongation artifacts in 3D reconstructions.
• Accurate 3D shape retrieval is challenging due to inherent biological morphological changes during development.

Purpose of the Study:

• To present a novel approach for correcting 3D artifacts in intensity-only tomographic acquisitions.
• To address the issue of axial elongation in 3D reconstructions of biological samples.

Main Methods:

• A network architecture combining volumetric and 3D finite object approaches was developed.
• The framework was applied to time-lapse images of mouse preimplantation embryos.

Main Results:

• The method successfully corrected axial elongation artifacts in 3D reconstructions.
• Spherical object shapes were recovered, demonstrating the effectiveness of the artifact correction.
• The approach was validated on developing mouse embryos from fertilization to the blastocyst stage.

Conclusions:

• This work presents a novel, unsupervised pipeline for correcting 3D tomographic reconstruction artifacts.
• The developed method shows potential for application to diverse biological samples and imaging conditions.
• It paves the way for improved 3D shape analysis in developmental biology and other fields.