An annotated human blastocyst dataset to benchmark deep learning architectures for in vitro fertilization

Area of Science:

• Reproductive Medicine
• Biotechnology
• Medical Imaging

Background:

• Assisted Reproduction Techniques (ART) are crucial for infertility treatment.
• Embryo selection is a key factor influencing ART success rates.
• Current methods rely on morphological assessment or time-lapse imaging, with AI showing promise but needing improvement.

Purpose of the Study:

• To develop a comprehensive dataset of static blastocyst images with expert annotations.
• To facilitate the training of deep learning models for predicting embryo developmental potential.
• To improve embryo selection accuracy and, consequently, clinical outcomes in in vitro fertilization (IVF).

Main Methods:

• Creation of a dataset comprising static blastocyst images.
• Inclusion of Gardner criteria annotations (a morphological scoring system).
• Collection of relevant clinical parameters associated with the embryos.

Main Results:

• A benchmark of human expert performance in annotating Gardner criteria is established.
• The dataset is designed for training deep learning models to predict Gardner's criteria.
• The dataset aims to predict clinical outcomes, including live birth, from static images.

Conclusions:

• The developed dataset is a valuable resource for advancing AI in embryo selection.
• Improved AI models can enhance the accuracy of predicting embryo viability.
• This research supports the development of more effective assisted reproduction technologies.