Development of a Machine Learning-Based Model for Accurate Detection and Classification of Polycystic Ovary Syndrome on Pelvic Ultrasound

Abstract

Polycystic ovary syndrome (PCOS) is a common endocrine disorder that disrupts reproductive function and hormonal balance. It primarily affects reproductive-aged women and leads to physical, metabolic, and emotional challenges affecting the quality of life. In this study, we develop a machine learning-based model to accurately identify PCOS pelvic ultrasound images from normal pelvic ultrasound images. By leveraging 1,932 pelvic ultrasound images from the Kaggle online platform (Google LLC, Mountain View, CA), we were able to create a model that accurately detected multiple small follicles in the ovaries and an increase in ovarian volume for PCOS pelvic ultrasound images from normal pelvic ultrasound images. Our developed model demonstrated a promising performance, achieving a precision value of 82.6% and a recall value of 100%, including a sensitivity and specificity of 100% each. The value of the overall accuracy proved to be 100% and the F1 score was calculated to be 0.905. As the results garnered from our study are promising, further validation studies are necessary to generalize the model's capabilities and incorporate other diagnostic factors of PCOS such as physical exams and lab values.