Early Diagnosis and Prediction of Polycystic Ovary Syndrome Using Machine Learning-Based Techniques: A Review

The prevalence of Polycystic Ovary Syndrome (PCOS), a complex endocrine condition that predominantly impacts women of reproductive age, has risen significantly in recent years. Excess concentrations of androgen and male hormone frequently constitute as contributing factors for PCOS. Hormonal deviations indeterminate menstrual cycles, and in-numerable tiny cysts on the female reproductive organs are characteristic features of PCOS. Beyond its detrimental impact on reproduction, PCOS has been associated to a wide range of metabolic disorders among which are obesity, dyslipidemia, insulin resistance, and a greater likelihood of cardiovascular disease and type 2 diabetes. However, if diagnosed early enough, PCOS-related challenges could be prevented or more effectively managed. It is vital to understand the difference between healthy and affected diagnostic observations in order to obtain an exhaustive understanding of the extent of the illness and its multifaceted detrimental impacts on women's reproductive health. Machine Learning (ML), considering its feature extraction capabilities, performs particularly well in detection among all the detection approaches. Various ML methods such as Random Forest, Support Vector Machine, Naive Bayes etc. are used in diagnosis of PCOS. An in-depth review of the techniques by which numerous machine learning-based algorithms have been implemented for identification of PCOS throughout the previous several decades is carried out and the approaches are explored in depth.