Article
MICROSCOPIC EMBRYO CLASSIFICATION USING AN INTEGRATED DEEP LEARNING FRAMEWORK
Embryo classification plays a crucial role in assisted reproductive technology (ART), especially in the context of in vitro fertilization (IVF). In India, IVF has witnessed rapid growth, with more than 1,500 clinics performing approximately 2.5 lakh cycles annually as of 2023, according to the Indian Society of Assisted Reproduction. Traditionally, embryo assessment is performed manually by embryologists using the Gardner grading system, which involves evaluating morphological characteristics such as blastocyst expansion, inner cell mass, and trophectoderm quality under a microscope. However, this method is inherently subjective, often influenced by the embryologist’s experience, fatigue, and inconsistent application of grading criteria. This subjectivity leads to significant inter- and intraobserver variability and limits the accuracy and scalability of embryo selection, contributing to stagnant IVF success rates of 30–40%. To address these challenges, a hybrid deep learning model combining Convolutional Neural Networks (CNN) with a Cat Boost Classifier (CBC) is proposed. This AI-driven approach aims to automate the classification of embryos into categories such as "normal" or "viable," thereby reducing human error and enhancing the predictive accuracy of implantation potential. CNNs are used to extract detailed features from microscopic images of embryos by resizing and normalizing pixel values, while the CBC performs efficient classification based on these features. The model not only improves consistency but also significantly boosts performance, achieving an accuracy of up to 99.06%. By enabling faster, data-driven, and objective decision-making, the proposed system overcomes the limitations of manual evaluation. It enhances embryo selection, increases implantation success rates, and offers a scalable solution for modern IVF practices in India
Full Text Attachment
