Accurate system for assessment and selection of human embryo after in vitro fertilization

Abstract

In Vitro Fertilization (IVF) is a complex medical procedure designed to assist individuals in overcoming infertility by facilitating the union of sperm and egg outside the human body. Not all embryos created during the IVF process have the same potential for successful implantation and development. Embryo grading allows embryologists to assess the quality of embryos based on specific criteria, such as cell division, symmetry, and the presence of fragmentation. By assigning grades, they can identify the most viable embryos with the highest likelihood of successful implantation. The conventional methods for embryo grading are time-consuming, labor-intensive, and prone to errors due to their subjective nature. Therefore, there is a pressing need for the development of an accurate and automated grading system to address these challenges. In this study, I propose a human embryo grading system based on machine learning. The proposed system utilizes embryo images as inputs and produces probabilities to predict the success or failure of each embryo. I utilize a convolutional neural network to develop the grading model and introduce a novel hierarchical feature extraction approach to enhance accuracy. Through extensive experiments, I demonstrate that the proposed method surpasses previous approaches with a significant performance margin.