Tumor cell phenotype recognition counting method based on cell fluorescence image

Abstract

The invention discloses a tumor cell phenotype recognition counting method based on a cell fluorescence image, and the method comprises the steps: collecting a cell nucleus fluorescence image and a plurality of cytoplasm fluorescence images, and converting the fluorescence images into a gray level image; determining a threshold correction factor of each grayscale image through a machine learning algorithm; correcting a binarization threshold value in a process of converting the grayscale image into the binary image according to a threshold value correction factor; generating a cytoplasm mask corresponding to a tumor cell phenotype through a cytoplasm binary image, generating a cell binary image in combination with a cell nucleus binary image, finally, recognizing tumor cells in the cell binary image, and counting the number of the tumor cells. Two methods based on rules and statistical machine learning are combined, various circulating tumor cell phenotypes can be fully automatically and flexibly recognized and counted under different image conditions, the recognition precision is high, counting is accurate, errors are small, and the recognition counting time is short.

Description

technical field [0001] The invention relates to the field of image analysis, in particular to a method for identifying and counting tumor cell phenotypes based on cell fluorescence images. Background technique [0002] Circulating tumor cells (CTCs) are tumor cells that shed from the original tumor and enter the blood circulation, and their counting has clinical oncological significance such as diagnosis, prognosis, and drug resistance prediction. Circulating tumor cells are often identified by immunofluorescence, typically using staining for nuclei, leukocytes, and tumor-specific protein markers, and are defined as nuclei positive for tumor-specific protein markers but negative for leukocyte markers. [0003] In addition, the heterogeneity of circulating tumor cells contributes to these cells having different phenotypes and expressing different tumor-specific proteins, for example, epithelial circulating tumor cells express cytokeratin and epithelial cell adhesion molecule ...

AI Technical Summary

Problems solved by technology

[0005] At present, neither of these two technical solutions can fully automatically and flexibly count different circulating tumor cell phenotypes under different image conditions

Although the rule-based scheme can write the rules for identifying different circulating tumor cell phenotypes into the system, it still needs to manually adjust the system parameters under different image conditions, so it cannot achieve a fully automated effect

Although the machine learning method can be fully automated, a large number of corresponding training set images are required to identify different circulating tumor cell phenotypes, and only known phenotypes can be trained and identified

Based on the above reasons, the existing technical solutions can only identify and count the most common epithelial circulating tumor cells, and other phenotypes such as mesenchymal, mixed and unknown circulating tumor cell phenotypes cannot be counted

Images