MEP-Net: A PIDNet-Based Model with Median-Enhanced Spatial-Channel Attention for Segmentation of Hepatocellular Carcinoma in CEUS Images

Purpose: Hepatocellular carcinoma (HCC) remains a major global health concern due to its high incidence and mortality. Contrast-enhanced ultrasound (CEUS) offers notable advantages in HCC diagnosis, including real-time imaging and non-invasiveness. However, challenges such as blurred lesion boundaries and noise interference in CEUS images significantly hinder the accuracy and robustness of automatic segmentation. Methods: To address these issues, we propose MEP-Net, an enhanced segmentation model based on PIDNet. MEP-Net incorporates a Median-Enhanced Spatial-Channel Attention (MECS) mechanism and an Efficient Channel Attention (ECA) module to better capture blurred areas and fine-grained structural details. We evaluate the model on a self-built CEUS dataset and the publicly available BUSI breast ultrasound dataset. Results: The results indicate that MEP-Net outperforms the baseline PIDNet by 1.96%, 1.96%, and 2.38% in Dice, MIoU, and Recall, respectively, on the CEUS dataset, and by 1.37%, 1.06%, and 2.41% on the BUSI dataset. In comparisons with eight mainstream segmentation methods, MEP-Net demonstrates superior boundary detection and small-lesion recovery, achieving leading overall performance. Ablation studies further confirm the complementary benefits of the MECS and ECA modules in boosting segmentation accuracy. Conclusion: The improvements of MEP-Net provide stronger support for CEUS image segmentation, which may have a positive impact on the early diagnosis and treatment of HCC.