High Precision Phase Recovery for Single Frame Fringe Pattern of Label-free Cells Detection Based on Deep Learning

As the basic unit of organism composition and life activity, the change of physiological state of cell is important to clinical disease prediction and diagnosis, especially blood diseases. In order to obtain the morphology of blood cells with abundant information content in 3D space without any biochemical or other complex processing for samples, this study proposed a transverse shear interference 3D imaging detection method for real-time dynamic label-free living cells based on deep learning. The phase extraction and recovery method of single red blood cell interference fringe image obtained by quantitative phase imaging system is carried out by Generating Antagonism Network (GAN). This method has a great improvement in efficiency and accuracy, it has a profound impact on the study of biological cells, and can be extended to the fields of cancer diagnosis and drug development in genomics.