Artificial Intelligence-Based Detection of Human Embryo Components for Assisted Reproduction by In Vitro...

Area of Science:

• Embryology
• Artificial Intelligence
• Medical Imaging

Background:

• Assisted reproductive technology (ART) assists in achieving pregnancy by addressing infertility.
• In vitro fertilization (IVF) involves combining sperm and eggs outside the body.
• Embryo morphology is critical for successful ART outcomes, with blastocyst development occurring in 3-5 days.

Purpose of the Study:

• To develop an automated method for analyzing blastocyst components to aid in embryo selection for IVF.
• To reduce the time and expertise required for manual microscopic analysis of embryos.
• To improve the accuracy and efficiency of identifying viable embryos for transfer.

Main Methods:

• Introduction of a deep learning-based embryo component segmentation network (ECS-Net).
• ECS-Net utilizes a shallow deep segmentation network with two distinct streams (base convolutional block and depth-wise separable convolutional block).
• Dense concatenation and dense skip paths are employed for powerful feature extraction during upsampling.

Main Results:

• The ECS-Net accurately segments key blastocyst components: trophectoderm, zona pellucida, blastocoel, and inner cell mass.
• Evaluation on a public microscopic blastocyst image dataset demonstrated the method's efficacy.
• The proposed ECS-Net achieved a mean Jaccard Index (Mean JI) of 85.93% for embryological analysis.

Conclusions:

• The developed ECS-Net provides an effective deep learning solution for automated embryological analysis.
• This AI tool can assist embryologists in selecting viable embryos, potentially increasing IVF success rates.
• The research highlights the potential of AI in reducing diagnostic burden and enhancing precision in assisted reproduction.