Liver cancer ranks among the most prevalent cancers globally, making early detection vital for effective treatment. Segmenting a liver and tumors from abdominall computed tomography (CT) is the first essential step for diagnosis. Despite the recent advances in computer aided diagnosis methods, the segmentation task remains challenging. The intensity of liver tissue is similar to that neighboring organ, making the process of boundary identification a challenging task. This work presents an automated system for classifying and identifying liver tumors in CT images through a hybrid segmentation approach which combines Neutrosophic Sets (NS), Watershed Transform, and Gaussian Mixture Model (GMM) to efficiently segment liver and tumors from CT images. The performance of the presented system is evaluated on publicly available datasets using standard metrics such as accuracy, sensitivity, and dice similarity coefficient. Experimental results indicate that our segmentation approach accurately identifies liver boundaries and reduces over/under segmentation. It proves to be effective, robust compared with to existing algorithms for liver and tumor segmentation in CT scan images.