A Stacked Ensemble Approach for IVF Outcome Prediction Using Gardner’s Blastocyst Grading System

Abstract

Infertility affects a significant proportion of couples globally, and in vitro fertilization (IVF) continues to be the most utilized and clinically effective form of assisted reproductive technology (ART) for addressing reproductive challenges. The success of human IVF treatment remains a challenging process, with a global success rate of less than 40% across all treatment cycles. Among critical factors that influence the success of an IVF cycle is selecting the optimal blastocyst using a standard grading system. This study presents a novel, fully automated stacked system designed to predict IVF outcomes, positive or negative pregnancy, using cutting-edge deep learning methodologies. The system incorporates three pretrained YOLOv8 models, each dedicated to grading specific blastocyst components (expansion, Inner Cell Mass (ICM), and Trophectoderm (TE)). The outputs of these models are integrated via logistic regression to deliver a robust prediction of IVF outcomes. A groundbreaking dataset of blastocyst images was gathered for this research, representing a diverse range of grading scenarios and outcomes, ensuring comprehensive training and evaluation of the system. Experimental results demonstrate that the proposed system achieves a high predictive accuracy of 83.3%, which represents a substantial leap in IVF success prediction. This study not only presents a new dataset but also establishes a new standard for automated blastocyst grading and outcome prediction, with the potential to improve clinical decision-making and patient outcomes in Assisted Reproductive Technology (ART).

Article Content

Keywords

Article Info