Oocyte Microscopic Image Fertilization Prediction based on First Polar Body Morphology using YOLOv8.

Area of Science:

• Reproductive Biology
• Medical Imaging
• Artificial Intelligence

Background:

• Oocyte development is critical for embryo viability and fertility in In Vitro Fertilization (IVF).
• The first polar body (PBI) is a potential factor influencing oocyte developmental competency, though its role remains debated.
• Accurate prediction of fertilization success and subsequent embryo development is essential in assisted reproductive technologies.

Purpose of the Study:

• To investigate the correlation between the first polar body (PBI) and the developmental competency of oocytes.
• To develop and validate an AI-based pipeline for predicting successful oocyte fertilization and development to the two-pronuclei (2PN) stage.
• To assess the utility of Convolutional Neural Networks (CNNs) in analyzing microscopic oocyte images for predictive purposes.

Main Methods:

• A pipeline utilizing You Only Look Once (YOLO) Convolutional Neural Networks (CNNs) was developed, comprising segmentation and classification models.
• Light microscopic images of metaphase II (MII) oocytes were captured, with the PBI segmented and cropped using a CNN.
• The cropped PBI images were then fed into a classification CNN to predict the likelihood of successful 2PN development post-intracytoplasmic sperm injection (ICSI).

Main Results:

• The YOLOv8l-cls model achieved high performance, with the best accuracy at 96.98%, sensitivity at 95.69%, and specificity at 98.28% on a dataset of 1006 images.
• The AI model successfully predicted the potential for 2PN development based solely on PBI image analysis, without requiring additional features.
• The results indicate a significant correlation between PBI characteristics and oocyte developmental potential.

Conclusions:

• The study demonstrates the efficacy of AI, specifically YOLO CNNs, in analyzing PBI morphology for predicting oocyte developmental competence.
• The developed AI pipeline offers a non-invasive method to assess oocyte quality and predict fertilization success in IVF.
• These findings contribute to a better understanding of the PBI's role in oocyte development and have potential implications for improving IVF outcomes.