Fig. 1: Utilizing deep learning Twin Networks to delineate embryonic developmental periods and reveal embryonic developmental rhythms.

A Flowchart of the entire experiment. Some of the material in (A) has been taken from Figdraw, with permission from Figdraw (license code: zzSSS0bb64). The image is created by the author, and there is no copyright or conflict of interest. B Schematic diagram of embryo age prediction. The test image (top panel) is compared with a sequence of reference images (middle panel) of known chronological order. The age of the test image corresponds to the age of the embryo image with the highest similarity (light blue dashed line) to the input data; relative (dark blue curve) similarity. The blue shading indicates the width of the peak. Some of the material in (B) has been taken from Figdraw, with permission from Figdraw (license code: zzSSS0bb64). The image is created by the author, and there is no copyright or conflict of interest. C Similarity plots of test embryos (top) versus reference images (bottom). Each test embryo was compared to three sets of reference images. The average of the cosine similarity with these reference image sets was plotted as a data point for each reference image time point. The box plots are based on distributions with similarity values above 0.8. The center of the graph represents the median, the box boundaries represent the upper and lower quartiles, the whiskers represent the 1.5-fold interquartile range, and the red dot represents the curve maximum. The three images in the figure are from a single embryo collection and represent three independent experiments. D, E Automated analysis of embryo development at different concentrations using Twin Networks.