Background <p>Endometrial cancer (EC) is a common gynecologic malignancy with rising incidence and significant molecular heterogeneity. This study aimed to develop an integrated prognostic model using pathomics features derived from histopathological images.</p> Methods <p>We retrospectively analyzed hematoxylin and eosin-stained whole slide images and clinical data from 511 EC patients in the TCGA database. Pathomics features were extracted using the same methodology as the reference study. Patients were randomly divided into training (<i>n</i> = 341) and validation (<i>n</i> = 170) cohorts at a 2:1 ratio. Under a leave-one-out cross-validation framework, features were selected using LASSO combined with random survival forest to construct a pathomics score. Differential gene expression and functional enrichment were analyzed and a nomogram integrating the pathomics score with clinical variables was developed and evaluated.</p> Results <p>The pathomics model demonstrated excellent prognostic prediction, with AUCs of 0.966, 0.724, and 0.918 in the training, validation, and whole cohorts for 5-year survival, respectively. The pathomics score showed significant associations with FIGO stage, grade, lymph node metastasis, and recurrence (<i>p</i> &lt; 0.05). Differential gene expression analysis revealed enrichment in EC-related pathways, MAPK signaling, estrogen signaling, and HIF-1 signaling pathways. Multivariable analysis confirmed FIGO stage, grade, lymph node metastasis, and pathomics score as independent prognostic factors. The nomogram incorporating these factors showed significantly improved in overall survival (all <i>p</i> &lt; 0.001 in the 3 cohorts) and predictive evaluation of AUCs (increases of 0.111, 0.132, and 0.118, respectively) with good calibration.</p> Conclusion <p>The proposed nomogram integrating pathomics and clinical factors provides accurate prognostic prediction for EC patients, offering a valuable tool for risk stratification and personalized management.</p>