The goal of the planned work is to build a system for assessing and predicting the potential
of gametes and embryos, as well as human reproductive capabilities, based on spectral data
obtained from the objects of investigation, followed by prospective validation of the
developed system. The study protocol consists of a retrospective and prospective stages. The
task of the retrospective stage is to study gametes, embryos, and endometrium using the
declared methods and build a machine learning model, determine the predictive capabilities of
the obtained models. The task of the prospective stage is to determine the practical
efficiency of applying models for making clinically significant decisions in infertility
treatment with IVF. Hypothesis of the study: at the moment, a large number of approaches and
protocols for deselecting and selecting embryos / gametes, assessing endometrial receptivity
has been proposed. Approaches related to deselection are mainly based on determining the
genetic constitution (aneuploidy) of the investigated object. However, there are no models
linking such testing results and the outcome of infertility treatment with clinically
significant effectiveness. There are many publications when, after transferring aneuploid
embryos, pregnancy develops with a healthy fetus. It is known that the concordance of
aneuploidy test results between the internal cell mass and trophoblasts is about 60%.
Moreover, when using PGT-a, the birth rate among women with a single available blastocyst is
reduced twice. Approaches related to selection, i.e. predicting a positive outcome of
treatment, are built on morphological, morphometric, metabolic, and gene expression
approaches. However, their effectiveness either has not been proven, or has (if it has)
relatively low predictive importance. This is due to the fact that, from the point of modern
views on reproductive biology, for the occurrence and development of successful pregnancy, it
is necessary to combine factors that belong to gametes, embryos, and the maternal organism.
Also, other undetectable technical or other circumstances may play a role in influencing the
chance of ongoing pregnancy. Therefore, for effective prediction of a positive outcome, it is
necessary to develop and apply complex models that take into account variables from different
sources, from all parties involved. However, there will always be additional variability,
caused by a series of unspecified or difficult to specify factors, which makes the task of
such prediction quite challenging. In this connection, predicting a negative outcome of
treatment (deselecting objects) seems more sensible, as it is entirely feasible for cases,
where the cause of the negative outcome is attributable to this object (for example, the
state of the embryo). This will not only optimize patient care protocols (for example, not to
transfer obviously incapable to implant embryos), but also determine the possible cause of
the negative outcome in each specific case, and in a population scale determine the share of
variability of the phenomenon (development of ongoing pregnancy), which may be related to a
specific object. The last one is necessary for adequate development and testing of new
therapeutic and diagnostic methods