Fluorescence microscope images of cells contain abundant phenotypic features,which are used to study the absorption and transportation of substances in cells,as well as chemical distribution and localization.These analyses require high-quality cell images.However,saturation artifacts will cause serious loss of phenotypic features,which will affect morphological analysis and certain classification experiments.From the perspective of data post-processing,a two-stage cell image inpainting model is proposed based on generative adversarial networks to solve the loss of phenotypic features caused by saturation artifacts.The model can restore large areas of missing phenotypic characteristics.The effectiveness and reliability of the restored image are validated through four groups of experiments.The results indicate that the restored results effectively fill in the missing phenotypic features and enhance the image quality for analysis.Classification experiments,serving as a representation of cell morphology analysis experiments,are conducted on both the original and restored cell images.It is proved that the image after restoring saturated artifacts can improve the experimental accuracy based on cell morphology analysis.
维普
万方数据