Predicting the Success of Blastocyst Implantation from Morphokinetic Parameters Estimated through CNNs and Sum of Absolute Differences

The process of In Vitro Fertilization deals nowadays with the challenge of selecting viable embryos with the highest probability of success in the implantation. In this topic, we present a computer-vision-based system to analyze the videos related to days of embryo development which automatically extracts morphokinetic features and estimates the success of implantation. A robust algorithm to detect the embryo in the culture image is proposed to avoid artifacts. Then, the ability of Convolutional Neural Networks (CNNs) for predicting the number of cells per frame is novelty combined with the Sum of Absolute Differences (SAD) signal in charge of capturing the amount of intensity changes during the whole video. With this hybrid proposal, we obtain an average accuracy of 93% in the detection of the number of cells per image, resulting in a precise and robust estimation of the morphokinetic parameters. With those features, we train a predictive model based on Random Forest classifier able to estimate the success in the implantation of a blastocyst with more than 60% of precision.