Digital staining of label-free cellular imaging results in optical diffraction tomography

Optical diffraction tomography (ODT) is an innovative three-dimensional label-free microscopic imaging technique that offers high spatial resolution and low phototoxicity, making it suitable for long-term super-resolution 3D imaging of live cells. However, label-free results cannot specifically distinguish cellular structures. The main challenges in using AI models for digital staining are the lack of datasets and model generalization issues. This report presents a method for quickly and accurately creating label-free cell datasets and utilizes these datasets to train AI models capable of identifying various cellular structures such as cell membranes, nuclear membranes, nucleoli, lipid droplets, and mitochondria. The models achieved an accuracy of over 85% on target test sets.