THE USE OF ARTIFICIAL INTELLIGENCE IN PREDICTING PREGNANCY IN IN VITRO FERTILIZATION: ANALYSIS AND CLASSIFICATION OF THE MAIN MORPHOLOGICAL, MORPHOKINETIC AND PATIENT VARIABLES

• Grant number: 23/13875-7
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): December 01, 2023
• Effective date (End): November 30, 2024
• Field of knowledge: Health Sciences - Medicine
• Principal Investigator: José Celso Rocha
• Grantee: Matheus Búbola Duarte
• Host Institution: Faculdade de Ciências e Letras (FCL-ASSIS). Universidade Estadual Paulista (UNESP). Campus de Assis. Assis , SP, Brazil

Abstract

Infertility affects approximately 15% of couples of reproductive age and can be attributed to various factors, such as lifestyle, genetic mutations and medical illnesses. Assisted reproduction has emerged as a solution to infertility, offering specific laboratory techniques adapted to the needs of each couple. However, the success rates of these techniques, such as in vitro fertilization, intracytoplasmic sperm injection and intrauterine insemination, generally do not exceed 40%.To increase the chances of success and reduce human error, assisted reproduction clinics have adopted innovative practices, such as pre-implantation genetic diagnosis and screening and the use of incubators with time-lapse technology, which capture images of embryonic cells and provide information on their morphology and morphokinetics. These innovations are accompanied by a growing interest in incorporating artificial intelligence.With the advance of technology, studies have shown the effectiveness of combining data collected by time-lapse, on morphology and morphokinetics, together with physiological information from the patient, applied in artificial intelligence models. Artificial intelligence uses techniques such as genetic algorithms and multi-layered artificial neural networks for training, allowing the creation of predictive models for gestational success. However, interpreting these results is challenging due to the non-parametric nature of the biological variables involved.In this context, this research project aims to use specific computational methods, such as calculating the weights of the connections between neurons, Garson's algorithm and partial derivatives, to identify the main variables involved in the artificial intelligence process that determine gestational success. Computer software will be developed on the Matlab platform that will use artificial intelligence techniques and the aforementioned methods to identify the most relevant variables in this process.