An Intelligent Analysis System For The Morphology Of Mouse Embryos After Implantation

Objective:1.Based on the mouse embryo morphological development score proposed by G.Van Maele-Fabry(1990),the database of early embryos after implantation was established,and the content richness and stability of the database were improved by improving the thrombectomy method and using Thunder image microcamera technology.2.Based on the YOLOv5 target recognition model and guided by the quantitative scoring standard of embryo development morphology,and using the self-built embryo image database as the learning sample,establish a set of objective,accurate and efficient automatic grading system in the early stage of mouse embryo development.Realize intelligent mouse embryo morphology score,information and scientific research,and automatic results analysis.Methods:1.Different versions of the rodent embryo morphology scoring system were compared and screened,the evaluation scope was comprehensive,the evaluation content was detailed,and the morphological development score of mouse embryo development was selected as the standard.2.Mice were housed,pooled and tested under appropriate environments.The pregnant mice were kept until the period required as guided by the scoring standards,and the embryos were extracted for backup.3.The collected mouse embryo samples from each period were collected from image samples one by one using Thunder image microscopy and stored in the tag image file format(.tiff).4.Using the labelme software based on Python environment,using the morphological development score of mouse embryos proposed by G.Van Maele-Fabry as the standard,the organs to be scored in the captured embryo images were annotated and scored.After completing the annotation,they were saved as a group of corresponding original images + scoring script files as the data set for AI learning.5.The embryo image dataset was enhanced by the Mosaic data to unify the picture parameters.The data set is divided into learning set and validation set,using YOLOv5 for learning set,according to the output layer error calculation loss,loss by backpropagation algorithm,through continuous optimization after modifying parameters,the validation set input has completed the learning process of YOLOv5 system target identification to verify the learning results.The confusion matrix of the learning objects was obtained,and the learning results were evaluated by various reference indicators(Precision,Recall,m AP,F1-score,etc.).Results:1.Different versions of the rodent embryo morphology scoring system were compared and screened to determine the morphological development score of the mouse embryo proposed by G.Van Maele-Fabry.2.A total of 1080 embryos were extracted by raising ICR mice,and 1595 embryo images were collected by embryo sample microphotography,including data from 6 time periods from E 0.25(6 h)and E 8.0 to E 10.0.The image contains 13 organ sites with the mouse embryo morphological development score proposed by G.Van Maele-Fabry(1990)as the standard reference for learning by the YOLOv5 target recognition system.3.YOLOv5 The target recognition system identified the 13 target organs in the mouse embryo images with the mean accuracy 0.866,0 mean recall 0.900,m AP mean0.884,F1-score mean 0.886,and the mean ratio 0.663 Io U.Combined with the learning results of the validation set and the manual annotation results,the recognition system has good learning and recognition results on the organ parts in the embryo image.Conclusion:1.The mouse embryo image database was successfully constructed using the morphological development score of mouse embryos proposed by G.Van Maele-Fabry.The mouse embryo image database built by using Thunder image microcamera technology has richer image content and database stability.2.By using the YOLOv5 target recognition model to learn the self-built mouse embryo image sample database,the recognition system has high recognition accuracy and accuracy,good repeatability and stability,which can more accurately reflect the development degree of the mouse embryo and the differentiation degree of the corresponding organs.An objective,accurate and efficient automated grading system was successfully established in the early stage of mouse embryo development.Realize intelligent morphological score of early embryo development after implantation,information and scientific research,and automatic results analysis.