Raman spectroscopy of follicular fluid and plasma with machine-learning algorithms for polycystic ovary syndrome screening.

Polycystic ovary syndrome (PCOS) is the main cause of anovulatory infertility and affects women throughout their lives. The specific diagnostic method is still under investigation. In the present study, we aimed to identify the metabolic tracks of the follicular fluid and plasma samples from women with PCOS by performing Raman spectroscopy with principal component analysis and spectral classification models. Follicular fluid and plasma samples obtained from 50 healthy (non-PCOS) and 50 PCOS women were collected and measured by Raman spectroscopy. Multivariate statistical methods and different machine-learning algorithms based on the Raman spectra were established to analyze the results. The principal component analysis of the Raman spectra showed differences in the follicular fluid between the non-PCOS and PCOS groups. The stacking classification models based on the k-nearest-neighbor, random forests and extreme gradient boosting algorithms yielded a higher accuracy of 89.32% by using follicular fluid than the accuracy of 74.78% obtained with plasma samples in classifying the spectra from the two groups. In this regard, PCOS may lead to the changes of metabolic profiles that can be detected by Raman spectroscopy. As a novel, rapid and affordable method, Raman spectroscopy combined with advanced machine-learning algorithms have potential to analyze and characterize patients with PCOS.