A Review On Machine Learning, Deep Learning, And Ensemble Transfer Learning Approaches with Hyperparameter Optimization for Polycystic Ovary Syndrome (PCOS) Detection from Ultrasound Sonography Images

Polycystic ovary syndrome (PCOS) is one of the most prevalent endocrine diseases among women of reproductive age. The problem of diagnosing this condition on the basis of ultrasound scanning pictures is complicated by various factors, including inter-patient variability, image noise, and follicle morphology irregularities. Researchers have been gradually shifting from conventional machine learning approaches to deep learning algorithms, transfer learning, and ensembled transfer learning models fine-tuned with the help of genetic algorithms, Bayesian Optimization, Optuna, and other metaheuristic techniques in order to increase the accuracy and generalization ability of PCOS diagnostics. The present paper provides a comprehensive overview of the latest research efforts aimed at creating automated PCOS detection systems. The main findings and performance of machine learning-based, convolutional neural network (CNN)-based, and transfer learning-based approaches are discussed in detail. Moreover, the paper highlights the efficiency of hybrid approaches, which combine optimized CNNs (EfficientNetB7, DenseNet201) and a Fuzzy Inference System applied through an image enhancement module, and advanced optimization methods, such as Genetic Algorithm. Particular attention is paid to the limitations of existing PCOS detection systems, including the lack of suitable data sets, and opportunities for future research are identified.