Capsule Networks have been developed as an alternative to Convolutional Neural Networks (CNN) for encapsulating spatial hierarchies in images or signals. The objective is to improve the precision of Polycystic Ovary Syndrome (PCOS) classification by sophisticated image processing methodologies. The proposed system uses the features of Capsule Networks to examine medical ultrasound images for PCOS classification while improving feature protection. It creates a reliable diagnostic model capable of accurately differentiating between healthy and PCOS. Capsule Networks provide more detailed evaluations of ovarian morphology by preserving the orientation and positioning information of characteristics. Three different capsule networks, such as Dynamic Routing CapsNet (DRCN), Expectation-Maximization (EM) Routing CapsNet (EMRCN), and Deep CapsNet (DCN), are analyzed for PCOS classification using more than 3000 images in the PCOS dataset. Results prove that the proposed Deep Capsule Network achieves better overall accuracy of 99.33 %, sensitivity of 99.27 %, and specificity of 99.4 % compared to other types of capsule networks. The combination of Capsule Networks with medical imaging procedures presents a promising framework for timely and precise diagnosis, thereby diminishing diagnostic delays and enhancing patient outcomes in gynaecological healthcare systems.

Capsule Networks Polycystic Ovary Syndrome Medical Image Processing Ultrasound Diagnosis Deep Learning Models