Advances in medical image segmentation have raised debate about the practical performance of the latest architectures and CNN-based approaches. While recent studies suggest that CNNs remain competitive, their performance in specific medical imaging tasks requires further validation. To address this, this study evaluates the performance of U-Net variants and STU-Net—a powerful scalable and transferable architecture, for hepatic tumor segmentation tasks. Our results on ATLAS dataset revealed that while traditional U-Net variants establish strong baseline performance, STU-Net achieved superior capabilities across various evaluation metrics, notably dice scores of \(95.76 \pm 0.99\%\) and \(68.30 \pm 2.13\%\) for liver and tumor segmentation respectively. These results validate its efficiency for such a challenging medical segmentation task.