Polycystic ovarian syndrome is an endocrine disorder that occurs in 4%–20% of all women in the world and is characterized by the presence of ovarian cysts, irregular menstrual cycles, hormone imbalance, hair thinning, acne and many more. It often results in metabolic and reproductive complications. More than 70% of the cases go undetected in the early stages as traditional diagnostic approaches involve time-consuming manual assessments of ovarian morphology which may lead to diagnostic inconsistencies due to variations in training, experience and judgement provided by different healthcare experts. In order to overcome the discussed challenges, this paper proposes a multi-modal framework that integrates Convolutional Neural Networks (CNNs) and K-Nearest Neighbors (KNN). CNNs are excellent in feature extraction from pre-processed ultrasonography images, while KNN classifies hormonal blood report data. This hybrid approach enhances the accuracy of diagnosis, reduces false positives, and ensures sensitivity in early diagnosis.