<p>Ovarian cancer remains one of the most fatal gynecologic malignancies, largely due to late-stage diagnosis and the limitations of current screening and imaging techniques. This review synthesizes recent advances in deep learning (DL) and Explainable Artificial Intelligence (XAI) for ovarian cancer detection across various modalities, including ultrasound, MRI, CT, histopathology, and multi-omics data. We highlight major DL trends, including CNN-, Transformer, and hybrid model-based approaches, and examine their performance, strengths, and underlying challenges such as data scarcity, model bias, and poor generalizability. The review also highlights the increasing importance of XAI techniques, including Grad-CAM, SHAP, and saliency methods, in enhancing transparency, clinician trust, and diagnostic decision support. Despite promising results, significant gaps persist, particularly in dataset standardization, multi-center validation, multimodal integration, and clinical deployment. We outline a future research roadmap prioritizing federated learning, hybrid multimodal pipelines, and robust XAI evaluation, while emphasizing the need for further validation and clinical translation before routine deployment.</p>