Non-Invasive PCOS Detection Using Saliva Based Hormone Analysis By Random Forest Classification

Abstract

Polycystic Ovary Syndrome (PCOS) is one of the most prevalent endocrine disorders affecting women of reproductive age worldwide. It is commonly associated with symptoms such as menstrual irregularities, infertility, metabolic abnormalities, and hormonal imbalance. Conventional diagnostic techniques primarily rely on blood tests and ultrasound imaging, which can be invasive, expensive, and often inaccessible in rural or low-resource healthcare settings. This study proposes a non-invasive and software-based method for the early detection of PCOS using hormone data derived from saliva samples and analyzed through a machine learning approach.Saliva provides a convenient and pain-free biological sample capable of reflecting hormonal fluctuations related to PCOS, including increased testosterone, elevated cortisol levels, and altered estrogen-progesterone balance. In this work, a Random Forest classification model is developed to analyze hormone-based features and predict PCOS status. The model is trained using either publicly available hormonal datasets or synthetically generated data designed to simulate realistic hormonal patterns.The proposed system includes stages such as data preprocessing, feature selection, model training, and performance evaluation. A user-friendly software interface allows users to input hormonal values and receive a prediction regarding PCOS likelihood. Experimental results demonstrate that the Random Forest model achieves high classification accuracy, indicating its potential as an effective decision-support tool. By providing a non-invasive and affordable diagnostic alternative, the proposed framework aims to improve accessibility to PCOS screening and increase awareness of women’s reproductive health, particularly in underserved communities.