OPTICAL SPECTROSCOPY RAMAN ASSOCIATED WITH ARTIFICIAL INTELLIGENCE ALGORITHMS TO PREDICT CONCEPTION AND GESTATIONAL STATUS IN BOVINE FEMALES

• Grant number: 24/16812-9
• Support Opportunities: Regular Research Grants
• Start date: May 01, 2025
• End date: April 30, 2028
• Field of knowledge: Agronomical Sciences - Veterinary Medicine - Animal Reproduction
• Principal Investigator: Luciano Andrade Silva
• Grantee: Luciano Andrade Silva
• Host Institution: Faculdade de Zootecnia e Engenharia de Alimentos (FZEA). Universidade de São Paulo (USP). Pirassununga , SP, Brazil
• Associated researchers: DANIEL YAEZU TIBA ; Guilherme Pugliesi ; Landulfo Silveira Junior ; Mariane Aparecida de Andrade Belezone ; Nara Regina Brandão Cônsolo ; Rafaella Zuliani Spalato ; Rogério Torres Seber ; Thiago da Cruz Canevari ; Willian Vaniel Alves dos Reis

Abstract

The global demand for food grows in parallel with the increase in population and stimulates the development of technologies to improve productivity. In beef cattle, technologies applied in recent decades to multiply genetic material and produce calves were essential for its production growth. However, the efficiency of many of these technologies can be improved. Fixed-time artificial insemination (FTAI), which has average conception rates of around 50%, is an example. Thus, the development of methodologies that can maximize FTAI efficiency could substantially contribute to greater calf production at lower costs. Early identification and selective use of females with greater conception potential and gestational diagnosis as early as possible are examples of actions that could substantially contribute to this. Techniques for clinical reproductive diagnoses that are highly accurate, fast and capable of being carried out in the field can be valuable for decision-making. With the rapid evolution of computational processing systems, artificial intelligence and machine learning are tools that allow the analysis of complex data, in large volumes and for diagnostic purposes. In this project, two studies will aim to develop methodologies to diagnose females with greater conception potential and for super early pregnancy diagnosis. Spectral data will be obtained by optical Raman spectroscopy in blood serum and cervical-vaginal mucus collected from bovine females subjected to the FTAI technique. Modifications in the composition of metabolites in these biological samples may be associated with different homeostatic statuses, mainly dependent on the modulation of metabolism by reproductive hormones. The samples will come from two animal experiments and two sample banks from previous experiments. After obtaining the spectral data, its dimensionality will be reduced and adjusted by principal component analysis and then subjected to analysis with machine learning algorithms for diagnostic determination. In Study 1 - Identification of animals with a higher probability of conception using optical Raman spectroscopy, in its first phase, machine learning algorithms will be developed to identify animals with a higher probability of conception, in addition to studying the repeatability of this characteristic. 200 samples of blood (serum) and cervical-vaginal mucus on days D-10, D-2 and D0 (day of AI), from experiment 1, will be analyzed by optical Raman spectroscopy to obtain spectral data. To evaluate the accuracy of the algorithms, the animal indicated as suitable for conception and which on days D22 and/or D30 are diagnosed as pregnant using ultrasound techniques will be considered a hit. The model will be tested and adjusted in phases 2 and 3 with samples from the sample bank and experiment 2. In Study 2 - Pregnancy diagnosis by optical Raman spectroscopy, in its first phase, machine learning algorithms will be developed for diagnosis of gestation, using samples from days D22 and D30 of the 200 females from experiment 1 to obtain spectral data. When evaluating the accuracy of the algorithms, the animal indicated as pregnant and which on days D22 and/or D30 were diagnosed as pregnant by ultrasound techniques will be considered a hit. The model will be tested in phases 2 and 3 with samples from the bank and experiment 2. In phase 3, the super early diagnosis of pregnancy during FTAI will also be investigated. In both studies, the accuracy of the results obtained between the biofluids analyzed will be compared. The aim of these two studies is to identify and describe patterns of spectral signatures (fingerprints) that allow the selection of females with a greater or lesser probability of conception and that allow for the diagnosis of early pregnancy quickly, accurately and at low cost. (AU)