A method and a device for predicting cellular autophagy phenotype

Abstract

The invention provides a method and a device for predicting cell autophagy phenotype. The method comprises the following steps of: collecting cell images, preprocessing the cell images, improving image quality and facilitating batch operation to obtain analysis images. According to the position of the cells in the analysis image, the analysis image is divided into a single cell image, so as to realize the accurate recognition of the single cell image. The single cell image is put into a trained convolution neural network model for calculation to obtain autophagy probability, and the cells corresponding to the single cell image whose autophagy probability is greater than a preset classification threshold are determined as autophagy cells, thereby avoiding manual extraction and identification, facilitating batch identification and prediction, and simultaneously realizing cell count.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a method and device for predicting cell autophagy phenotype. Background technique [0002] Autophagy is a mechanism for cells to remove harmful proteins and recycle substances to maintain cellular homeostasis. Under normal physiological conditions, autophagy relies on lysosomes (or vacuoles) to wrap damaged organelles, protein aggregates, long-lived proteins, etc., and transport them to lysosomes or vacuoles for degradation. According to the different autophagy functions and substrates, autophagy can be divided into macroautophagy, microautophagy and chaperone-mediated autophagy. We usually refer to autophagy as macroautophagy. Under the conditions of starvation, hypoxia and low energy, a double-layered vesicle structure will be formed in the cytoplasm, which will continuously wrap the cytoplasmic components and finally close to form autophagosomes, which will then fuse with ly...

AI Technical Summary

Problems solved by technology

However, manual analysis of fluorescence images, search for regions of interest, and extraction of image features are time-consuming and laborious. Although some intelligent learning systems have emerged, none of them can achieve the purpose of batch intelligent prediction and identification of autophagy phenotypes.

Images