Analysis of follicular fluid and serum markers of oxidative stress in women with unexplained infertility by Raman and machine learning methods

Abstract

AbstractOocytes are surrounded by a fluid called follicular fluid, which provides an essential microenvironment for developing oocytes in human fertility. Various molecules exist in antral follicles, including proteins, steroid hormones, polysaccharides, metabolites, reactive oxygen species, and antioxidants. Oxidative stress is involved in the etiology of defective oocyte development or poor oocyte and embryo quality. Raman spectroscopy, a noninvasive method, can be used for biological diagnostics and direct chemical identification of follicular fluid. Therefore, we measured the oxidative index of follicular fluids and then attempted Raman spectroscopy on the follicular fluids combined with machine learning techniques to identify, detect, and quantify follicular fluid of unexplained infertility‐diagnosed women as a safe and effective tool to use as adjacent for clinical studies. This was a retrospective study set in an academic hospital where the patients were selected from an unexplained infertility‐diagnosed population in the in vitro fertilization (IVF) center. Raman spectra of 128 follicular fluid samples (n = 63 control; and 65 unexplained infertility) were obtained. To profile Raman‐based results of follicular fluid, oxidative load measurements, multivariate analysis, correlation tests, and six machine learning methods were used. Raman bands associated with oxidative load and amide III and lipids differed significantly. Classification using stacks of Raman signals was applied by random forest, C5.0 decision tree algorithm, k‐nearest neighbors, deep neural networks, support vector machine, and XGBoost trees algorithms achieved an overall accuracy of 92.04% to 99.17% in assigned correctly. Group has an oxidative load in their follicle fluids consistent with clinical results and biochemical measurements and performing testing based on Raman spectra validated by kNN clustering and SVM object vector separation machine learning methods. The study suggests that Raman spectroscopy can detect changes in follicle fluid in unexplained infertility.