Research And Application Of Automatic Recognition Algorithms For Embryo Images

In vitro fertilization and embryo transfer(IVF-ET)technology provides an effective solution for the global medical problem of human infertility.Evaluating embryo quality and optimizing transplantation by embryo morphology is a necessary method to improve success rate and clinical pregnancy rate.However,the current manual evaluation method still has problems such as long time-consuming,subjective,and poor consistency.How to evaluate the continuous state of embryo development through automatic recognition technology is a hot topic in the field of assisted reproduction,but the accuracy of the existing automatic recognition algorithm for embryo images is relatively low.Therefore,it is of great significance to study the automatic recognition technology of embryo images and improve the recognition accuracy of embryo images.In this paper,an automatic recognition method for embryo images is proposed for the problem that the existing automatic recognition algorithm of embryo images is not accurate.The method optimizes the image preprocessing algorithm,improves the automatic recognition algorithm of embryo blastomere,and constructs the automatic positioning,number and critical event prediction methods.The main research contents are as follows:1.Aiming at the important influence of embryo image preprocessing results on image recognition accuracy,an embryo image preprocessing method was proposed.The method calculates the conversion factor to achieve the amplification compensation,and combines the mask method to remove the interference information.Based on the analysis of the seven edge detection algorithms,the EDContours with the F value of 0.62 is selected as the edge detection algorithm for the embryo image.2.In view of the loss of edge detail,petri dish and debris interference in blastomere recognition,this paper proposes an improved algorithm based on EDCircles.The EDPF edge detection step in EDCircles is replaced by EDContours,which solves the problem of edge detail loss in blastomere recognition.The threshold constraint method is applied to solve the problem of dish and debris interference.Through comparative analysis,the P,R and F values of the singleembryo and multi-embryo detection are better than the other five methods.This indicates that the algorithm has good accuracy and versatility,and can realize rapid detection and precise positioning of embryo blastomeres.3.Aiming at the feedback and prediction of continuous frame of embryo cleavage stage,a new automatic recognition method based on Markov chain was constructed.The method is based on the number constraint problem between successive frames of embryo images,combined with the event probability characteristics of Markov chain,generates label space for each frame of image,and determines the optimal label of the entire sequence to achieve automatic embryo positioning,number,and key event prediction.