A Novel Approach to Evaluating the Synergistic Effect of Curcumin and Cisplatin on Breast Cancer Cells: AgNORs Staining Based on Machine Learning

Cisplatin, an antineoplastic drug commonly used to treat many solid tumors, has serious side effects that limit its clinical use. Therefore, combination therapy with curcumin may be a good option to increase efficacy. The number, size, and distribution of Argyrophilic Nucleolar Organizing Regions (AgNORs) in the nucleus are useful in tumor detection and prognosis; however, since manually determining AgNORs is time-consuming and error-prone, a method that allows for accurate and rapid determination is crucial. In this study, we aimed to evaluate the antiproliferative effects of curcumin and cisplatin combinations on breast cancer cells by analyzing AgNOR staining using machine learning methods. Therefore, we determined the IC50 values of cisplatin and curcumin on MCF-7 breast cancer cells using the MTT test and calculated Combination Index (CI) values in combined applications. Approximately 100 AgNOR-stained interphase nuclei per group were analyzed using radiomic feature extraction and machine learning-based classification to quantify treatment-induced nuclear texture changes. Combinations of cisplatin and curcumin at different concentrations reduced cell viability to 41.41% (± 8.59%) and 20.52% (± 4.24%), respectively, with CI values less than 1. Consistent with the MTT assay, AgNOR-based radiomic features revealed significant changes in nuclear tissue and heterogeneity, demonstrating suppression of nucleolar activity and proliferation at the microstructural level. These findings demonstrate that radiomic AgNOR assay numerically mirrors the antiproliferative effects measured by conventional viability assays. Classification using machine learning models showed the highest discrimination level (AUC = 0.988 and accuracy = 0.970) between the control and the cisplatin-curcumin combination group. To the best of our knowledge, this study is the first available research on the morphometric analysis of interphase AgNOR proteins using machine learning applications to evaluate the efficacy of different treatment agents in a cancer cell line.