An Intersection of Artificial Intelligence and Healthcare: A Focus on Polycystic Ovary Syndrome Diagnosis

Polycystic ovary syndrome (PCOS) is a prevalent condition affecting women worldwide, with early detection being essential for improving treatment outcomes and preventing complications such as infertility. Diagnosing PCOS is challenging due to its heterogeneous nature and overlapping symptoms, including irregular menstrual cycles, obesity, acne, and hirsutism. Artificial intelligence (AI), particularly machine learning (ML), has significantly advanced the detection and diagnosis of medical conditions, including PCOS. However, the application of AI in PCOS diagnosis has not been comprehensively assessed. This study conducts a bibliometric review and analysis of AI applications in PCOS diagnosis to map research trends, identify gaps, and highlight potential directions for future research. This bibliometric review initially retrieved 4846 articles related to the application of AI, particularly ML, in the diagnosis of PCOS from various academic search engines and digital libraries. Following a rigorous screening process, 114 articles published between 2010 and 2023 were selected for final analysis. These studies were evaluated based on core methodological components, including data sources, feature engineering, classification techniques, validation strategies, preprocessing methods, segmentation, and cyst detection. Additionally, the review examined evolving trends in AI methodologies, with particular emphasis on the adoption of ML, deep learning, and swarm intelligence in PCOS diagnosis. The review uncovered several key findings. First, although AI, particularly ML, has been employed in PCOS detection, the body of research remains limited, especially regarding the application of deep learning techniques. Furthermore, the analysis revealed that existing preprocessing methods for ovarian ultrasound images are inadequate and lack practical robustness. In addition, only three studies employed swarm intelligence for feature selection, indicating a significant research gap in this domain. While electronic health records (EHRs) emerged as the most commonly used data source, image-based datasets from platforms such as Kaggle were often of low quality, characterized by duplicated images, poorly executed data augmentation, and insufficient sample sizes. Furthermore, no studies addressed cyst localization, and segmentation efforts were minimal, thereby pointing to critical underexplored areas. Finally, techniques such as federated learning for data privacy and explainable AI approaches remain largely underutilized in PCOS research, further emphasizing the need for more comprehensive and technically advanced investigations. This bibliometric analysis emphasizes the need for further research in several critical areas related to PCOS diagnosis, including cyst localization, image segmentation, attention mechanisms, and data preprocessing. The findings indicate that the application of artificial intelligence, particularly deep learning and swarm intelligence, remains limited in this field. Additionally, enhancing the quality of image-based datasets is essential, as current resources often exhibit issues such as poor augmentation and image redundancy. Addressing data privacy concerns through the adoption of federated learning techniques is also vital for enabling broader clinical use. Although explainable AI methods offer significant potential for improving interpretability and clinical trust, they have not yet been widely implemented. Consequently, future research should focus on these areas to strengthen the effectiveness and reliability of AI-based approaches for the detection and diagnosis of PCOS.