<![CDATA[Polycystic Ovary Syndrome (PCOS) is a complex hormonal disorder that significantly impacts the reproductive and metabolic health of women. A primary challenge in PCOS analysis is the reliance on subjective manual observations of ultrasound (USG) images, which are prone to inconsistency due to the lack of objective standards. This study proposes an automated, unsupervised framework to stratify the morphological characteristics of PCOS ultrasound images without predefined clinical labels. The method integrates feature extraction using the VGG16 deep learning architecture with K-Means clustering. The structural integrity of the computational clusters was quantitatively evaluated, achieving a Silhouette Score of 0.4248 and a Davies-Bouldin Index (DBI) of 0.9175, indicating the successful formation of distinct morphological groupings. Furthermore, an evaluation using a Support Vector Machine (SVM) to assess the internal consistency and linear separability of the extracted features yielded a test accuracy of 97.54%. The results demonstrate that the integration of VGG16 and K-Means effectively partitions PCOS ultrasound images into objective, computationally stable morphological groups. This approach is designed to function as a standardized clinical decision support tool, mitigating visual subjectivity and providing a measurable baseline for women's reproductive health management.]]>
Copyrights © 2026
