Inception and LightGBM-based Model for Poly-Cystic Ovary Syndrome Detection with CNN in Ultra-sound Images

Nowadays, a common hormonal condition that affects women of reproductive age and frequently results in infertility is Poly-Cystic ovary syndrome. Recognizing the Poly-Cystic ovary syndrome early is essential for symptom management and mitigating associated health risks. The traditional methods are used for diagnosis, encompassing high androgen levels, anovulation, and Poly-Cystic ovaries observed via ultra-sound. However, the challenging aspect of manually assessing Poly-Cystic ovaries observed via ultra-sound through follicle counting and thickness determination by medical professionals persists. In addition to addressing patient symptoms, clinicians must rely on a combination of biochemical and clinical indicators, lacking standardized diagnostic procedures. To alleviate these diagnostic challenges, this paper presents a hybrid Poly-Cystic Ovary Syndrome Detection model using a convolutional neural network comprising an ensemble inception model and a light gradient-boosting machine on ultra-sound Images. It significantly improves the detection of the Poly-Cystic ovaries observed and achieves an accuracy of 98.4%, surpassing InceptionV3’s 96.3%. This study substantiates this performance gain through comprehensive quantitative effectiveness assessment metrics. This study proposes a promising model for accurate and efficient Poly-Cystic ovary syndrome diagnosis, aiding clinicians in providing timely and targeted care to affected individuals.