Lens-free holographic microscopic particle characterization method based on convolutional neural network

Abstract

The invention discloses a lens-free holographic microscopic particle characterization method based on a convolutional neural network. The method comprises the following steps: S1, firstly, acquiring adark field image, then, acquiring a bright field image uniformly illuminated by a light source; S3, placing a sample above a sensor, acquiring microscopic images of the sample of different refractiveindexes, and marking corresponding refractive index of each image; S3, performing flat-field correction for all the holographic microscopic images; SS4, calculating a center of all particles in the images, and cutting images of each particle; S5, cleaning all the cut images, randomly dividing the images into a training set, a verification set and a test set; taking the training set as input of the convolutional neural network, training a classification network, verifying an effect training parameter on the verification set, finally, testing a classification effect on the test set, wherein a classification label corresponding to the particle is a refractive index characterization result of the particle. The method provided by the invention can perform quick, convenient and accurate characterization for biological samples under a large field of view.

Description

technical field [0001] The invention belongs to the field of microscopic images, in particular to a method for characterizing lensless holographic microscopic particles based on a convolutional neural network. Background technique [0002] Lensless holographic microscopy has emerged as a new imaging technique in recent years. In order to obtain high resolution, traditional optical microscopes must use magnification objective lenses and eyepieces to observe tiny biological images. The lensless holographic microscope completely abandons the optical lens and directly samples the light passing through the object. As a digital holographic technology, light is captured by the photosensitive array of the sensor, and image information is displayed through photoelectric conversion, and subsequent image processing can be conveniently performed. In addition, the compact structure of lensless holographic microscopes and the consistent size of the field of view with the imaging sensor ...

AI Technical Summary

Problems solved by technology

Thanks to the speed and efficiency of the convolutional neural network, the autofocus algorithm in the field of lensless microscopy can be quickly and equivalently implemented by the deep convolutional neural network, and the tracking of colloidal particles can also be completed by the deep convolutional neural network. However, the existing characterization of various particles still uses the above-mentioned heuristic algorithm to perform complex and time-consuming iterations, which is far from meeting the needs

Images