Purpose The ability to detect, monitor, and precisely time the morphokinetic stages of human embryos plays a critical role in assessing their viability and potential for successful implantation. In this context, the development and utilization of accurate and accessible tools for analysing embryo development are needed. This work introduces a highly accurate, machine learning model designed to predict 16 morphokinetic stages of pre-implantation human development, which is a significant improvement over existing models. This provides a robust tool for researchers and clinicians to use to automate the prediction of morphokinetic stage, allowing standardisation and reducing subjectivity between clinics. Method A computer vision model was built on a public dataset for embryo Morphokinetic stage detection containing approximately 273,438 labelled images based on Embryoscope/+ embryo images. The dataset was split 70/10/20 into training/validation/test sets. Two different deep learning architectures were trained and tested, one using efficient net V2 and the other using efficient-net V2 with the addition of post-fertilization time as input. A new postprocessing algorithm was developed to reduce the noise in predictions of the deep learning model and detect the exact time of each morphokinetic stage change. Results The proposed model reached an overall test accuracy of 87% across 17 morphokinetic stages on an independent test set. If only considering plus or minus one developmental stage, the accuracy rises to 97.1%. Conclusion The proposed model shows state-of-the-art performance (17% accuracy improvement compared to the best models on the same dataset) to detect morphokinetic stages in static embryo images as well as detecting the exact moment of stage change in a complete time-lapse video.
