Lightweight U-Net Based Liver Tumor Segmentation Using CT Scans

Accurate liver segmentation is a critical component of medical image processing tasks, including treatment planning and disease diagnosis. Traditional methods, such as manual segmentation, are time-consuming, computationally intensive, and prone to variability. To address these limitations, this study proposes a lightweight 2D U-Net architecture for liver segmentation. The model is trained and evaluated using the LiTS17 dataset, which provides high-resolution, annotated CT scans for liver tumor segmentation. To address the issue of class imbalance, a weighted segmentation loss function based on the Dice Coefficient is incorporated. Experimental results demonstrate that the proposed model outperforms existing segmentation approaches, offering a more efficient and effective solution for liver tumor segmentation.