Learning-based multi-spectral meta-optics for extended depth of field computational microscopy

Fluorescence microscopy is essential for analyzing complex subcellular structures, yet multispectral imaging often suffers from geometric distortions caused by chromatic aberrations and from the inherently shallow depth of field (DoF) of high-NA systems. As a result, capturing fully in-focus volumetric information typically requires depth scanning, which is further hindered by scattering in thick biological specimens. Existing extended-depth-of-field (EDoF) solutions help mitigate these issues but remain limited in broadband, multispectral performance. In this presentation, we introduce an enhanced design framework that integrates end-to-end optimized full-color meta-optics, experimentally validated in a 4f fluorescence microscopy system. Our approach jointly learns a meta-optic placed at the Fourier plane and a physics-guided reconstruction algorithm, enabling simultaneous correction of defocus blur and chromatic aberrations. This co-designed optical–computational system delivers sharp, distortion-free extended-depth-of-field imaging across a broad spectral range, offering significant improvements for robust multispectral fluorescence microscopy.