Ovarian cancer remains a significant challenge in the realm of oncology due to its heterogeneous nature and often late-stage diagnosis. Addressing this, we propose a comprehensive approach to ovarian cancer research focusing on subtype classification and outlier detection using both image and numeric data.  For the image data, we utilize a combination of Convolutional Neural Networks (CNN) and MobileNet, DenseNet architectures trained on a curated dataset comprising various ovarian cancer subtypes, including High-Grade Serous Carcinoma (HGSC), Clear-Cell Ovarian Carcinoma (CC), Endometrioid (EC), Low-Grade Serous Carcinoma (LGSC), and Mucinous Carcinoma (MC). Upon uploading an image, our model will predict the subtype with high accuracy, facilitating precise diagnosis and treatment planning also add on explainability. In parallel, for numeric data analysis, we leverage a dataset focusing on Polycystic Ovary Syndrome (PCOS) and employ outlier detection techniques. Our approach involves applying algorithms such as Logistic Regression, Decision Trees, Random Forest, and XGBoost to identify anomalies indicative of PCOS presence. This dual-pronged strategy enables a holistic understanding of ovarian cancer, combining advanced image analysis with numeric insights for comprehensive patient care. The proposed backend implementation in Python ensures robust model development and deployment, while the frontend built using HTML, CSS, and JavaScript provides an intuitive user interface for seamless interaction. Together, our methodology aims to advance ovarian cancer research, enhance subtype classification accuracy, and improve outlier detection for better patient outcomes.

 Keywords: Ovarian cancer, Subtype classification, Outlier detection, Convolutional Neural Networks (CNN), MobileNet, DenseNet, High-Grade Serous Carcinoma (HGSC), explainability.