Published Paper Details:

This study presents a robust deep learning framework for the segmentation and classification of liver tumors from CT scan images, utilizing the publicly available 3Dircadb dataset. A comprehensive methodology is adopted, encompassing data preprocessing, augmentation, and dynamic generation to enhance training efficiency. Central to the approach is the ResUNet architecture, which integrates the strengths of U-Net and residual learning to accurately delineate liver tumors. The model is trained using the Adam optimizer with Dice coefficient loss, and its performance is evaluated using key metrics such as pixel accuracy, true positive rate, and Dice score. The system achieved a Dice coefficient of 0.89 and a true positive accuracy of 99.68%, demonstrating its high reliability in medical image segmentation. Comparative analysis with existing models highlights a notable improvement in accuracy and robustness. This research offers a promising direction for clinical diagnostics and computer-aided treatment planning in hepatology.