Liver cancer remains a critical global health challenge characterized by high incidence and mortality rates. In clinical practice, the accurate segmentation of liver lesions is often hindered by their morphological complexity and a heavy reliance on time-consuming, expert-driven manual annotations. To address these limitations, this study proposes an automated framework for liver lesion mask generation based on a multi-objective, multi-threshold formulation. The approach simultaneously optimizes four objective functions: Otsu, Minimum Cross-Entropy (MCE), Minimum Error Thresholding (MET), and Tsallis Entropy. Experimental evaluations demonstrate that the proposed methodology achieves a DICE coefficient of 0.751, outperforming several supervised segmentation models within the Liver Tumor Segmentation Benchmark (LiTS). These results highlight the method's potential as an efficient, lightweight, and unsupervised alternative for enhancing diagnostic workflows in medical imaging.