Convex optimization-based mesoscopic 3D reconstruction method for fluorescence-free light fields

Currently, the microscopic 3D observation primarily relies on specific fluorescence staining, which will cause irreversible damage to samples and presents issues such as high cost and phototoxicity. A novel method for mesoscopic high-resolution 3D reconstruction without fluorescence was proposed, known as RGB-MesoLFM. First, a fluorescence-free labeled white-light mesoscopic circuit was constructed, where the RGB three-band light field data were decoupled. Subsequently, based on fluctuating optics point spread function, the point impulse response of the system at any depth within the three bands was computed. Finally, an objective function was developed using convex optimization theory, considering the RGB three-band light field data and point impulse response as input for 3D deconvolution iteration, which can reconstruct the slice image of the high-resolution target object at any depth. The 3D reconstruction experiment was based on the light field sampling results of egg embryo slices, and could obtain slice images in the range of 0~−20 µm with a sequence interval of 4 µm. Compared with the traditional fluorescence reconstruction method, the proposed method does not require fluorescence, but uses white light for imaging. The reconstructed peak signal to noise ratio (PSNR) value of arbitrary depth slices is about 10% higher than that of traditional methods.