Deep Learning for Preoperative MRI-Based Endometrial Cancer Staging Prediction

Purpose:Endometrial carcinoma ranks among the most common malignancies of the female reproductive system. Accurate early-stage staging is essential for devising appropriate treatment plans and assessing patient prognosis. This study aims to enhance diagnostic precision by overcoming the limitations of traditional imaging methods and existing deep learning models. Methods:To address challenges such as dependency on physician expertise, inefficiency, and deficiencies in feature transmission, boundary detail restoration, and multi-scale feature integration, we propose a novel architecture termed GCMF-UNet. Furthermore, for classification tasks, we introduce MSFA-Net, which integrates a ResNet-18 backbone with a multi-scale feature aggregation module, squeeze-and-excitation (SE) attention, and a Swin Transformer for global contextual modeling. Results:Experimental results indicate that GCMF-UNet surpasses the traditional U-Net by 4.1%–5.5% in key evaluation metrics, including Accuracy and Recall. In classification performance, MSFA-Net achieves a 2.4%–3.7% improvement over baseline models across multiple quantitative indicators, demonstrating enhanced capability in identifying critical lesion regions. Conclusion:The proposed GCMF-UNet and MSFA-Net architectures effectively mitigate limitations of conventional diagnostic and deep learning approaches, offering more accurate lesion segmentation and classification. These advancements provide a promising foundation for improving automated diagnosis and staging of endometrial carcinoma.