Deep learning for predicting stem cell efficiency for use in beta cell differentiation.

Recent clinical trial data show curative potential of cell therapy for diabetes, however the cells required are a bottleneck. Cell differentiation exhibits substantial variability, even among clones of stem cells generated from the same patient. Human experts struggle to see the difference between highly- and lowly-efficient cell clones early. We therefore propose an image-based deep learning model to guide the selection of the most efficient clones. We apply different deep learning models to learn the morphological differences between good and bad stem cell clones and classify them based on phase-contrast imaging. To gain insight into the learned features, we use layer-wise relevance propagation, and Fourier-based frequency analysis. Using an EfficientNet-V2-S model, we obtain a novel early prediction for the outcome of the differentiation process from patient-derived stem cells to [Formula: see text] -cells using imaging. Clone level accuracy is 96.7 % at 53 hours after start of differentiation. The analysis of learned features shows that the structure of the cell population is an important predictive feature. This study is a proof-of-concept that deep learning combined with label-free imaging can be highly predictive and guide selection of stem cell clones, thereby reducing cost of [Formula: see text] -cell production.