Efficient Liver and Tumor Segmentation Using a Compact Residual Network and Contrast-Enhanced Pre-Processing Pipeline

Accurate liver and tumor segmentation in CT images is vital for diagnosis and treatment planning. This study presents SegResNet_2335, a lightweight 3D residual network optimized for volumetric segmentation. Combined with a tailored pre-processing pipeline-including voxel spacing resampling, CT window adjustment, CLAHE, and z-score normalization-the proposed framework achieves strong and consistent segmentation performance. On the LiTS test set, average Dice Similarity Coefficients (DSCs) reached 0.956 for liver and 0.754 for tumor segmentation. Using the finalized sp1.5_win_clahe_z preprocessing configuration, evaluation on the independent 3D-IRCADb-01 dataset yielded a liver DSC of 0.847 and a tumor DSC of 0.706, demonstrating robust cross-dataset generalization. The model architecture contains only 1.5 million parameters and supports rapid inference (∼1.8 s per scan), making it suitable for real-time and resource-constrained clinical deployment. The complete implementation is publicly available to support reproducibility.