New Trends in Ovarian Cancer Diagnosis Using Deep Learning: A Systematic Review

Ovarian cancer (OC) is one of the most common types of cancer in women. Surgery and chemotherapy are still the most common forms of treatment; however, their success depends on lots of factors describing the type of cancer, the size, shape, and its origin, and thus early and accurate detection could bring lots of benefits to increasing survival rate by applying custom/ personalized and effective treatment. This is why many researchers aim to obtain accurate computer-aided diagnosis (CAD) systems to assist in the early detection and diagnosis of such diseases. The current paper presents a systematic review of new trends in designing different deep learning-based intelligent systems for accurate OC detection and diagnosis. The paper presents the advantages of using deep learning approaches for OC diagnosis, the most used methods, and datasets. It performs a detailed analysis concerning the most preferred, effective, and accurate architecture. 95 articles published in journals and impact conferences were investigated between 2018 and 2024, focusing on 2021 and 2024. All included studies are indexed in PubMed, Scopus, or ISI Web of Science. The authors of the analyzed articles used private databases (the majority) and public databases. Neural networks and combined DL systems recently used for ovarian tumor detection, classification, and segmentation were investigated. The performances obtained by the most relevant systems were also presented. Finally, a comparison with other reviews in the field highlighted the advantages of this article.