Deep Learning Assisted Plasmonic Dark-Field Microscopy for Super-Resolution Label-Free Imaging

Advanced Quantitative Phase Microscopy Achieved with Spatial Multiplexing and a Metasurface

Quantitative optical phase information provides an alternative method to observe biomedical properties, where conventional phase imaging fails. Phase retrieval typically requires multiple intensity measurements and iterative computations to ensure uniqueness and robustness against detection noise. To increase the measurement speed, we propose a single-shot quantitative phase imaging method with metasurface optics that can be conveniently integrated into conventional imaging systems with minimal modification. The improvement of the measurement speed is simultaneously made possible by combining deep learning with the transport-of-intensity equation. As a proof-of-concept, we demonstrate phase retrieval on both calibrated phase objects and biological specimens by using an imaging system integrated with our metasurface. When combined with the matched neural network, the system yields result with errors as low as 5% and increased space-bandwidth-product. A multitude of commercial applications can benefit from the compactness and rapid implementation of our proposed method.