All‐Dielectric Meta‐Microlens‐Array Confocal Fluorescence Microscopy

Abstract Acquisition time and optical sectioning capability are critical factors in fluorescence imaging. Confocal microscopy is a vital optical imaging method to inherently observe volumetric tissues with fine optical sectioning capability; however, point‐by‐point scanning is time‐consuming. Metasurfaces, a type of flat optics utilizing nano‐scale structures, provide diverse functionalities and extensive flexibility in controlling light wavefronts. Here, meta‐microlens‐array (meta‐MLA) for multifocal confocal fluorescence microscopy to enhance acquisition speed is introduced, reduce photo‐bleaching, and improve energy efficiency while remaining compatible with existing commercial scanning configurations. Point spread function (PSF) in the meta‐MLA confocal lateral and axial directions has been evaluated. Fast optically sectioned images of various samples, including pollen grains and biological tissue phantoms, are performed. Image quality is further enhanced by the Richardson–Lucy (RL) deconvolution method with total variation (TV). The trade‐off between spatial resolution and acquisition speed is overcome using deep neural network models, comparing performance metrics with a conventional confocal microscope. The combination of meta‐MLA confocal and deep learning with superior image quality and fast acquisition will likely extend the clinical applications of miniaturized optical imaging.