Quantification of follicles in human ovarian tissue using image processing software and trained artificial...

Area of Science:

• Reproductive biology
• Medical imaging
• Artificial intelligence

Background:

• Cancer treatments can cause premature ovarian insufficiency (POI), leading to infertility and endocrine dysfunction.
• Ovarian tissue transplantation is a promising fertility restoration method, but its success depends on follicle density.
• Manual analysis of follicle density is labor-intensive and time-consuming.

Purpose of the Study:

• To develop and validate a fully automated artificial intelligence (AI) method for quantifying follicle density in human ovarian tissue.
• To improve the efficiency and accuracy of selecting ovarian tissue for transplantation to treat POI.

Main Methods:

• Developed a three-stage AI pipeline for follicle identification and object classification.
• Trained the AI using manually annotated histologic sections of ovarian tissue from 12 donors (ages 16-37).
• Compared AI performance against two manual operators analyzing whole slide images.

Main Results:

• The AI pipeline correctly identified 80% of follicles.
• The automated method was 33% faster than manual annotation.
• AI analysis demonstrated high potential for accurate follicle counting.

Conclusions:

• AI-driven automation offers a faster and efficient alternative to manual follicle density assessment.
• This technology can aid in selecting optimal ovarian grafts for improved longevity in POI treatment.
• Automated follicle quantification is vital for advancing fertility preservation strategies.