机器学习辅助图像识别红细胞抗氧化研究OA北大核心CSTPCD

Cell morphology,a pristine biological feature,provides intrinsic information on cell physiological or pathological conditions in a different manner than biochemical indicators.Image recognition methods based on artificial intelligence(AI)are helpful in analyzing speed and accuracy of cell morphology.In this study,an oxidative damage prediction model for cell morphology image segmentation,identification and counting was established,and the results were highly consistent with flow cytometric cell counts.Further,the established model was used to study the dynamic process of antioxidation in erythrocyte.The results showed that the ratio of normal morphological erythrocytes obtained by machine learning-assisted image recognition showed consistent trends with the classical biochemical indicators,which indicated that the model could effectively predict the degree of oxidative damage of red blood cells.Moreover,the method could also intuitively observe the real-time changes in erythrocyte morphology without staining or cell fragmentation.The model was expected to be expanded for rapid indicator screening,especially on cell morphology,cell viability and other environmental toxicology applications.