Exploring Embryo Development at the Morula Stage - an AI-based Approach to Determine Whether to Use or Discard an Embryo

Assisted reproductive technology addresses infertility by combining sperm and egg in-vitro and implanting the resultant embryo into the female for a healthy birth. The success of ‘assisted reproduction’ relies on selecting the most viable embryo, a task manually carried out by embryologists. Embryologists assess the viability of an embryo within a cohort by examining the morphological development of embryo cleavage stages. However, this selection process is subjective and resource intensive. This study advances the field of ART by employing artificial intelligence to assess embryos at the morula stage. The morula stage is an essential yet less explored cleavage stage in embryo development. This approach attempts to bridge the existing gap in understanding the morphological structure of the morula and potentially transform assisted reproductive processes by improving the embryo selection criteria. Specifically, training a video classifier to analyze the morphological development of embryos from the beginning to the end of the morula stage and predict the embryo’s fate-whether to transfer, cryopreserve, or discard it. The classifiers efficiently predicted ’discarded’ videos and frequently confused between ‘transferred’ with ‘cryopreserved’ videos. The classifier understanding of the morphological characteristics at the morula stage was investigated by using the Grad-CAM explainable AI technique, providing embryologists deeper insights into embryo development at this stage. The GradCAM heatmaps showed distinct morula development patterns for predicted outcome: ‘cryopreserve’ and ‘discard’ videos, while ‘transfer’ videos lacked relevant morphology cues.