High-resolution three-dimensional diffraction tomography based on confocal microscopy

Abstract

Confocal microscopy, with its exceptional optical sectioning capabilities and multimodal imaging capabilities, provides critical tools for morphological observation and quantitative analysis. As a label-free technique, confocal transmission mode offers rich information on cellular 3D structures; however, its optical transfer function suffers from an inherent “missing cone” problem, which causes severe degradation of axial resolution and generates stretching artifacts, thereby limiting detailed 3D quantitative analysis. The Gerchberg-Papoulis (GP) iterative algorithm can compensate for missing spectral information through frequency-domain extrapolation; however, its reconstruction performance is highly dependent on the accuracy of the spatial support domain, and transmission images themselves struggle to precisely extract the three-dimensional boundaries of complex biological samples. To address this issue, this paper proposes a 3D reconstruction method that integrates dual-modality imaging with the GP algorithm: leveraging the superior optical sectioning properties and 3D resolution of confocal fluorescence imaging, high-fidelity 3D support domains are obtained by specifically labeling the cell membrane; this support domain is then introduced as a core constraint into the GP algorithm to perform frequency-domain extrapolation and reconstruction of transmission images within the same field of view. Experimental results on cell models demonstrate that this method effectively suppresses axial stretching artifacts, significantly improves axial resolution and structural fidelity, and provides a robust and efficient new computational imaging strategy for the analysis of fine cellular structures.