Deep learning for quantifying spatial patterning and formation process of early differentiated human‐induced pluripotent stem cells with micropattern images.

Micropatterning is reliable method for quantifying pluripotency of human‐induced pluripotent stem cells (hiPSCs) that differentiate to form a spatial pattern of sorted, ordered and nonoverlapped three germ layers on the micropattern. In this study, we propose a deep learning method to quantify spatial patterning of the germ layers in the early differentiation stage of hiPSCs using micropattern images. We propose decoding and encoding U‐net structures learning labelled Hoechst (DNA‐stained) hiPSC regions with corresponding Hoechst and bright‐field micropattern images to segment hiPSCs on Hoechst or bright‐field images. We also propose a U‐net structure to extract extraembryonic regions on a micropattern, and an algorithm to compares intensities of the fluorescence images staining respective germ‐layer cells and extract their regions. The proposed method thus can quantify the pluripotency of a hiPSC line with spatial patterning including cell numbers, areas and distributions of germ‐layer and extraembryonic cells on a micropattern, and reveal the formation process of hiPSCs and germ layers in the early differentiation stage by segmenting live‐cell bright‐field images. In our assay, the cell‐number accuracy achieved 86% and 85%, and the cell region accuracy 89% and 81% for segmenting Hoechst and bright‐field micropattern images, respectively. Applications to micropattern images of multiple hiPSC lines, micropattern sizes, groups of markers, living and fixed cells show the proposed method can be expected to be a useful protocol and tool to quantify pluripotency of a new hiPSC line before providing it to the scientific community. LAY DESCRIPTION: Human‐induced pluripotent stem cells (hiPSCs) are an ideal resource for patient specific regenerative medicine. Pluripotency of a hiPSC line has to be analysed before us. Differentiated hiPSCs on a micropattern have been used for the analysis on which hiPSCs form segregated, sorted and radially ordered three germ layers. This study uses DNA‐stained fluorescence microscopy images for segmentation of the differentiated hiPSCs, together with fluorescence images staining respective ectoderm, mesoderm and endoderm to classify a segmented hiPSC as any of the three germ layers. However, cells on a DNA‐stained fluorescence micropattern image are low‐contrast to background and dense that are not easily segmented by image‐processing methods. Meanwhile, regions of fluorescent pixels on the germ‐layer fluorescence images overlap, although the hiPSCs differentiating on a micropattern expand in a two‐dimensional, nonoverlapped way at the early differentiation stage. We propose a U‐net structure with flexible input micropattern sizes and with designed loss functions to identify the dense differentiated hiPSCs to improve segmentation accuracy. We then use an intensity comparison calculation that compares fluorescence intensities of all the corresponding pixels on the germ‐layer fluorescence images inside a segmented hiPSC to classify the cell into a nonoverlapped region of any germ layer. Then, distribution (spatial pattering) of a germ layer on the micropattern are quantified using numbers, areas and distances to the micropattern centre of cells inside the germ layer. The U‐net is also trained by the labelled cells on DNA‐stained fluorescence images with their corresponding bright‐field images. This enables the hiPSC segmentation on live‐cell time‐lapse bright‐field micropattern images to quantify spatial patterning changes of the germ layers over time or monitoring of their formation process. [ABSTRACT FROM AUTHOR]