Quantitative and qualitative metrics of tumor stroma in predicting ovarian cancer outcomes and expansion of its study with AI-based tools

Abstract

Epithelial ovarian cancer remains one of the deadliest gynecologic malignancies, with late-stage diagnosis, high recurrence rates, and resistance to platinum-based chemotherapy contributing to poor survival outcomes. Central to the effective management of ovarian cancer is the thorough evaluation of diagnostic and prognostic indicators. Critical determinants encompass the extent of the tumor; its stage and grade; and level of the circulating biomarker, CA-125. Additional tumor cell-centric factors such as BRCA1/2 mutation status, homologous recombination deficiency, and folate receptor-alpha (FRα) protein levels inform initial treatment and maintenance strategies. Unfortunately, these markers alone cannot fully predict outcomes or significantly improve survival rates. This review emphasizes the body of data suggesting that both quantitative and qualitative metrics of tumor stroma play a crucial role in the prognosis and outcomes of epithelial ovarian cancer. We examine quantitative and qualitative metrics such as stromal proportion, tumor density, stiffness, and texture. We explore how artificial intelligence (AI) tools advance the measurement of these parameters, offering unprecedented opportunities to integrate stromal biomarkers into clinical decision-making. By synthesizing emerging evidence, we propose a framework for leveraging stromal properties-individually and in combination-as novel prognostic indicators to improve outcomes for patients with ovarian cancer.

Keywords: MT: Regular Issue; artificial intelligence; epithelial ovarian cancer; outcomes; prognostic factors; tumor stroma; tumor texture; tumor-stromal proportion.

Graphical abstract

Figure 1

Characterization of tumor-stroma-centric metrics and summary of their relevant studies in epithelial ovarian cancer (A) Depicts the tumor-stroma metrics discussed throughout the review. (B) Summary of pertinent studies discussed in the review. Abbreviations: AUC, area under the curve; AUROC, area under the receiver operating characteristic curve; CT, computed topography; EMT, epithelial-mesenchymal transition; HR, hazard ratio; IC₅₀, concentration of carboplatin to inhibit 50% of cell growth; PFS, progression-free survival; OS, overall survival; PET, positron emission tomography; TSP, tumor stroma proportion; 2D, two-dimensional; 3D, three-dimensional. ^aStudy used artificial intelligence techniques. ^bData expressed as mean ± standard deviation. Created in BioRender, Madill, M. (2025); https://BioRender.com/dqcrkdj.

Figure 2

Application of artificial intelligence in ovarian cancer research and clinical practice Categorization of various areas of practice and research where artificial intelligence has been applied and benefits and limitations of its application. Created in BioRender, Madill, M. (2025); https://BioRender.com/9cvsfq2.