Automatic Classification of Red Blood Cell Morphology Based on Quantitative Phase Imaging

Classification of the morphology red blood cells (RBCs) plays an extremely important role in evaluating quality long-term stored blood, as RBC storage lesions such transformation discocytes to echinocytes and then spherocytes may cause adverse clinical effects. Most segmentation classification methods, limited by interference staining procedures poor details, are based on traditional bright field microscopy. In present study, quantitative phase imaging (QPI) technology was combined with deep learning for automatic morphology. QPI can be used observe unstained RBCs high spatial resolution information. information, boundary curvature is reduce inadequate preliminary screening three shapes RBCs. The model accuracy 97.3% stacked sparse autoencoder plus Softmax classifier. Compared convolutional neural network, developed method showed a lower misclassification rate less processing time, especially more discocytes. This has potential applications automatically real-time diagnosis RBC-related diseases.