A hybrid cell type identification method based on fine-grained recognition

Abstract

The present invention specifically relates to a method for identifying mixed cell types based on fine-grained recognition, comprising the following steps: pre-establishing a fine-grained recognition convolutional neural network model and a cell image database, the cell image database includes images of mixed cells, mixed cells The image is an image including multiple types of cells; S1, collect images of mixed cells; S2, input the images of mixed cells into the fine-grained recognition convolutional neural network model, and obtain a heat map of cell types; S3, compare the images of mixed cells Perform thresholding to obtain a binary image of the cell area; S4, combine the binary image of the cell area and the heat map of the cell type to obtain the identification result of the cell type. The invention accurately identifies cell types according to the specificity of cell morphological characteristics, and avoids the disadvantages of long time-consuming and cumbersome process of traditional cell type identification methods. The model can learn fine-grained cell morphological features, and identify cell types through texture and other information, with high recognition accuracy and robustness.

Description

technical field [0001] The invention relates to the fields of biomedical image processing and machine learning, in particular to a method for identifying mixed cell types. Background technique [0002] In biomedical experiments, cell lines are often misidentified or cross-contaminated. The use of misidentified or cross-contaminated cell lines will lead to serious consequences such as irreproducible experimental results, wrong research conclusions, and clinical cell therapy disasters. It also wastes A lot of manpower, energy, money, etc. The traditional cell line identification method uses the method of comparing the DNA information of the cell sample with the cell bank loci to determine the type of the cell line and whether it is cross-contaminated, which is costly and time-consuming. [0003] Recently, deep convolutional neural networks have achieved great success in many vision tasks. Compared with traditional machine learning methods, convolutional neural networks do no...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the defects of the prior art, the present invention provides a method for identification of mixed cell types that is easy to operate and has accurate results. Accurate identification of cell types is performed according to the specificity of cell morphological characteristics, avoiding the long time-consuming traditional cell type identification methods. , The disadvantages of cumbersome process

Images