The Automated Calculation of Human Ovarian Reserve

We report on the use of AI techniques – including image analysis, feature recognition, machine learning and constraint satisfaction – applied to the problem of determining the population of primordial follicles in a human ovary. We highlight the importance of accurate population data for studies that aim to predict the age of menopause in healthy women, and for studies that quantify damage to ovarian tissue as a result of radioor chemotherapy. We describe existing methods used to estimate primordial follicle populations, and review the strengths and weaknesses of these approaches. We present a novel methodology for counting each and every primordial follicle in an ovary, together with details of our implementation and empirical evaluation.