A method using artificial neural networks to morphologically assess mouse blastocyst quality

Area of Science:

• Embryology
• Artificial Intelligence
• Biotechnology

Background:

• Embryo morphological classification is crucial for laboratory techniques and assisted reproduction.
• Current classification methods are subjective, relying on embryologist expertise.
• Developing objective, automated blastocyst quality assessment is needed.

Purpose of the Study:

• To develop software for classifying blastocyst morphological quality using digital images.
• To create an objective and reproducible method for embryo assessment.
• To reduce subjectivity in embryo quality evaluation.

Main Methods:

• Utilized artificial neural networks (ANNs) as a machine learning technique.
• Developed software analyzing visual variables from images and biological embryo features.
• Created a graphical user interface (GUI) for user-friendly operation.

Main Results:

• Achieved a final system accuracy of 95% after the training process.
• The GUI allows quality assessment based on a trained ANN.
• The software provides an objective assessment of embryo morphology.

Conclusions:

• The developed method demonstrates high potential for applicability, adaptable to various species including humans and cattle.
• Offers an objective assessment, minimizing embryologist bias.
• Ensures high reproducibility across different samples, clinics, and laboratories, serving as an alternative practice for embryo morphology assessment.