Increasing a Microscope’s Effective Field of View via Overlapped Imaging and Machine Learning

Abstract

We demonstrate a multi-lenses microscopic imaging system that records over- lapping fields-of-view for high-efficiency automated specimen analysis. We show both in simulation and experiment how our system can achieve accurate target object detection on overlapped images.

Large depth-of-field ultra-compact microscope by progressive optimization and deep learning

“…In addition to potential use in behaving animals, the integrated microscope is a multipurpose instrument for various applications, including flow cytometry 9 , air quality monitoring 57 , and cancer screening 58 . For in vitro applications, the integrated microscope has the potential to achieve even higher throughput on a large scale through massively parallel strategies 12 and is capable of being equipped with other instrumentation such as incubators thanks to the miniaturized sizes 9 . With upcoming GPU advances in improved speed, efficiency, and reduced size, integrated microscopes in intelligent platforms seem likely to facilitate the emerging paradigm of mobile analysis, screening, and diagnostic evaluations.…”

“…Emergent advances in sophisticated optical design try to circumvent these restrictions through complex lens configuration 6 and multi-view information acquisition 7,8 , which achieve remarkable results in table-level laboratory equipment but the bulkiness is even more problematic.Miniaturized integration is a pivotal advance that facilitates low-cost production and typically leads to improved performance and broad applications in telecommunications, computing, and genomics 9 . Recently, miniaturized microscope has achieved breakthroughs in multiple aspects including neural recording in freely behaving mouse 9,10 , high-throughput screening 11,12 , and flow cytometry 13,14 . Further with computational enhancement, the extended DOF (EDOF) that provides robustness over rough surfaces of 3D samples 15 can be achieved together with corrected color and enhanced resolution [16][17][18] .…”