Super-resolution fluorescence imaging techniques have emerged as a pivotal solution to examine sub-cellular processes. Among these techniques, Localized Plasmonic Structured Illumination Microscopy (LPSIM) has shown remarkable potential to achieve three-fold resolution improvement and video imaging speed by leveraging plasmonic nanoantenna arrays. However, the conventional design process for these arrays is hindered by time-consuming trial-and-error simulations and limited design degree of freedom. We introduce a hybrid inverse design framework that combines deep learning and genetic algorithm-based optimization to enhance the capabilities of LPSIM. Our approach yields optimized nanoantenna arrays that demonstrate superior reconstruction quality, offering robustness against noise, and requiring fewer measurements. The proposed method not only streamlines the efficient design process for LPSIM nanoantenna arrays, but also opens new avenues for exploring and optimizing plasmonic nanostructures for a wide range of applications beyond imaging.
|