Predicting Response to Treatment and Survival in Advanced Ovarian Cancer Using Machine Learning and Radiomics: A Systematic Review

Simple Summary The study initially presents an overview of the literature on studies using machine learning in ovarian cancer. The most common algorithms are neural networks. Since this topic is exceedingly vast, we then focus on studies reporting the use of machine learning and radiomics to predict treatment response, surgical outcome, and survival. For the prediction of the response to neoadjuvant chemotherapy, most studies had a performance around 80%, which is higher than usual statistics; these studies extracted data from clinical parameters and radiology, and may help select patients for surgery. In patients undergoing surgery, the models described in the study had a performances also over 80%. Furthermore, the highest performing model was in a platinum-resistant setting with an AUC > 95%, which can be particularly clinically useful to select patients for novel targeted therapies such as mirvetuxumab. We also summarize clinical biomarkers that may be used to predict survival. Thus, the role of ML and radiomics in ovarian cancer is very promising to improve the prediction of treatment outcomes and to assist in decision making for optimal treatment modality to ultimately improve survival. Future prospective studies are needed to confirm these findings; however, our study presents the first overview of the available literature.