Prediction of gestational success, using the variables of morphology and morphokinetics of the human embryo and the patient, using the Artificial Intelligence technique, through deep learning and multilayer perceptron.

• Grant number: 23/16156-1
• Support Opportunities: Regular Research Grants
• Start date: September 01, 2024
• End date: August 31, 2026
• Field of knowledge: Health Sciences - Medicine
• Principal Investigator: José Celso Rocha
• Grantee: José Celso Rocha
• Host Institution: Faculdade de Ciências e Letras (FCL-ASSIS). Universidade Estadual Paulista (UNESP). Campus de Assis. Assis , SP, Brazil
• Associated researchers: Marcelo Fábio Gouveia Nogueira
• Associated scholarship(s): 24/23599-0 - Automated extraction of mathematical variables representing the morphology of in vitro produced equine embryos, BP.IC

Abstract

Infertility, nowadays, is being considered a public health problem, affecting around 15% of the world's population. The increase in conditions harmful to fertility is related to factors such as: obesity, alcohol use, drug use, sexually transmitted diseases, among others. In addition to these facts, in today's world, competition is quite fierce and many couples have chosen to leave pregnancy on the back burner, largely due to achieving better socioeconomic conditions and then opting for motherhood/fatherhood. In this scenario, the assisted reproduction technique has been in high demand worldwide, whether through in vitro fertilization or intracytoplasmic sperm injection. It is estimated that since the birth of Louise Brown in 1978, 8 million people have been born with the help of assisted reproduction techniques around the world. In Europe, between 1997 and 2018, around 2 million births took place using assisted reproduction techniques and in Brazil, between 2020 and 2022, around 120,000 cycles of ovarian stimulation were performed and approximately 66,000 embryos were transferred. However, assisted reproduction techniques still have a low pregnancy success rate, around 25% to 40%, largely due to the subjectivity involved in choosing the embryo to be transferred. Currently, with incubators equipped with the time-lapse system, which allows the continuous cultivation of embryos, monitoring their development in a non-invasive way and within a controlled environment, which also provides the acquisition of photomicrographs that are taken at a predetermined time interval, which when evaluated jointly and continuously demonstrate morphological and morphokinetic characteristics of the embryo, this subjectivity has decreased, but also promoted the possibility of other techniques being used. Among these techniques, Artificial Intelligence has stood out as a non-invasive technique with great potential for use in assisted reproduction techniques, providing a higher accuracy rate in predicting gestational success. The application of the Artificial Intelligence technique, in the most diverse forms, artificial neural networks, multilayer perceptron, machine learning, deep learning, genetic algorithms or convolutional neural networks, among others, can provide a thorough analysis of both the image of the embryo and of cleavage times, in addition to being able to associate the patient's physiological characteristics with these variables, which would be impossible to obtain using the traditional methods that currently exist. In this sense, the present research project aims to develop artificial intelligence software, on the Matlab platform, which, having as input variable the morphological and morphokinetic data of the embryo in addition to the patient's data, uses artificial intelligence techniques, in particular artificial neural networks multilayer perceptron, genetic algorithms and convolutional neural networks, together or separately, to predict gestational success and thus greatly reduce the subjectivity that currently exists in choosing the embryo to be transferred to the patient. And so that the developed software can be used in everyday assisted reproduction clinics, a graphical interface will be developed so that it is easy to use and understand the results by embryologists and doctors working in assisted reproduction. (AU)