Radiomics-based prognostic model for progesterone resistance in endometrial cancer: insights into extracellular...

Progesterone resistance in fertility-preserving patients with endometrial cancer (EC) remains a significant challenge, and radiomics has not yet been used to predict progestin therapy in these cases.

In this study, we constructed a radiomics model to predict progesterone resistant for fertility preservation patients. Distribution of clinical features have significant differences in the high and low risk of progesterone-resistant subgroups, which we call predicting-sensitive (PS) versus predicting-resistant (PR) subgroups. The radiomics model achieved high predictive accuracy with an area under the receiver operating characteristic curve (AUC) of 0.841 in the training cohort. We further validate this model in the validation and whole cohorts. As a result, the AUCs were 0.873 and 0.852, respectively. The identified key biological pathways include cellular response to external stimulus, collagen metabolic processes, and extracellular matrix (ECM) remodeling. PS was strongly linked to higher type III collagen content and changes in ECM stiffness, which were reflected in altered tumor microenvironment dynamics. Finally, we confirmed that the PR subgroup exhibited increased cellular stiffness by microfluidic device. Exogenous supplementation of COL III enhanced the sensitivity of EC cells to progesterone and reduced cell stiffness.

Our radiomics model provides a promising, noninvasive tool for predicting progesterone resistance in EC. This approach paves the way for personalized therapeutic strategies for fertility-preserving patients.