SE-VGGNet: Liver tumor detection using squeeze excited VGG network

Liver tumors, including metastatic cancer and cholangiocarcinoma, pose a significant threat to global health, necessitating accurate and timely detection. This paper presents a novel deep learning model, SE-VGGNet, for liver tumor classification, which innovatively combines Squeeze-and-Excitation (SE) blocks with a VGGNet backbone. This integration enhances feature extraction by allowing the model to recalibrate channel-wise feature responses, focusing more effectively on the most relevant image regions. SE-VGGNet was evaluated on a liver tumor dataset, achieving a classification accuracy of 99.09%, precision of 98.15%, and recall of 100%, significantly outperforming traditional CNN and VGG-16 models. The novelty of this approach lies in its ability to achieve high performance even with small datasets, addressing a critical challenge in medical imaging. Additionally, this method offers the potential for broader application to other medical imaging tasks, highlighting its adaptability and impact on early cancer diagnosis and clinical decision-making.