Quality evaluation of induced pluripotent stem cell colonies by fusing multi-source features

Guanghui Yue,Jinqi Liao,Yongjun Wang,Liangge He,Tianfu Wang,Guangqian Zhou,Baiying Lei

DOI: https://doi.org/10.1016/j.cmpb.2021.106235

Abstract:Background and objective: Induced pluripotent stem cells (iPSCs) have great potential as the basis of regenerative medicine. In this paper, we propose an automatic quality evaluation model based on multi-source feature ensemble learning to divide the iPSC colonies into three categories: good, medium and bad. Methods: First, we obtained iPSCs samples using a Sendai virus reprogramming method. Second, we collected the bright field-images of iPSC colonies and processed them with adaptive gamma transform and data enhancement. The evaluation for the iPSC colony quality was further verified with living cell fluorescent staining, currently accepted as the optimal biological method. Third, multi-source features were extracted using three deep convolutional neural networks (DCNNs) and four traditional feature descriptors. Finally, we utilized a support vector machine (SVM) to perform classification. Before feeding into the SVM, the features were processed by principal component analysis algorithm to save computational cost and training time. Results: Experimental results on the collected iPSC dataset (46,500 images) show that the proposed method could obtain 95.55% classification accuracy. Conclusions: Our study could provide a method to efficiently and quickly judge the biological quality of a single iPSC colony or populations and facilitate the large-scale iPSC manufacturing.