基于深度学习的虚拟相衬成像方法

Deep Learning-Based Virtual Phase Contrast Imaging Method

对于传统的显微镜来说,其相衬成像模式需要通过配备专用光阑、聚光镜或者在物镜上添加嵌件来实现,这就增加了相衬显微成像的难度和成本。鉴于此,本文提出了一种基于深度学习算法的虚拟相衬成像方法,只需要采用一台普通的光学明场显微镜获取细胞明场图像,然后使用深度学习方法,就可以将明场图像转换成相衬图像。将虚拟相衬图像与显微镜获取的标准相衬图像进行对比,对比结果证明了这种虚拟相衬成像方法的有效性,为低成本相衬显微成像提供了范例。

For conventional microscopes,the phase contrast imaging mode requires the configuration of special diaphragms,condenser or the addition of inserts to the objective lens,which increases the difficulty and cost of phase contrast microscopic imaging.Therefore,a method of virtual phase contrast imaging based on deep learning algorithm is proposed.Only an ordinary optical bright field microscope is required to acquire cellular bright field images,and then the bright field images are converted to phase contrast images on a computer using a deep learning method.We compare the virtual phase contrast images with the standard phase contrast images acquired by the microscope.The results demonstrate the effectiveness of this virtual phase contrast imaging method,which provides an example of low-cost phase contrast microscopic imaging.