Artificial intelligence techniques for assisting the decision of making or postponing the embryo transfer

Human infertility is an important social problem, particularly stressed in the recent years, and intensive research is being done in the field of therapies such as Assisted Reproduction Technologies (ART) to cope with it in the most difficult situations. The success of ART depends on several important contributing factors: the success in ovarian stimulation, fertilization and embryo implantation. This last step is the most frustrating one, failing in most transfer of embryos in uterus. Together with real evaluation of oocyte and embryo inner quality one of the most relevant aspects of ART and reproduction in general is, consequently, the assessment of endometrial development and receptivity for embryo implantation, in order to predict the probability of pregnancy. The present study is aimed to assess if endometrial, subendometrial, uterine blood flow and other parameters measured using non invasive techniques can be used to predict a pregnancy: the proposed system does not limit its analysis to the statistical significance of the single measure considered but combines them together in a data mining approach. Our experiments performed on a dataset of 98 cycles of intracytoplasmic sperm injection (ICSI) show that it is possible to measure in a non invasive way a set of features from a patient, for predicting a pregnancy and also assisting the decision of physicians and biologists to make or postpone the embryo transfer.