Fertility-sparing treatments are becoming increasingly important for patients worldwide with atypical endometrial hyperplasia (AEH) and early endometrial cancer (EC). The primary aim of these treatments is to achieve complete regression (CR). Accurately predicting CR and applying effective treatment strategies throughout the process are crucial for successful fertility preservation in these patients. In our cohort, we included 155 patients and selected six radiomic features strongly correlated with CR from over 1500 MRI-based features. Tn area under the curve (AUC) value of MR score achieved up to 0.871. We then incorporated MR scores and clinical factors into a nomogram, demonstrating a strong predictive performance. With the introduction of MR score, the AUC value reached as high as 0.871, indicating a significant improvement in net benefit. In summary, we introduced an innovative MR score and developed a nomogram that significantly improves predictive accuracy for the likelihood of CR in patients seeking fertility preservation. Moving forward, we plan to establish a multicenter patient cohort to further refine this model and support its broader implementation in clinical practice following thorough evaluation.