This survey paper comprehensively reviews the recent advancements in liver tumor detection and segmentation utilizing deep learning models. The escalating incidence of liver tumors necessitates accurate and efficient diagnostic tools, and deep learning has emerged as a promising solution. The survey systematically categorizes and evaluates diverse deep learning approaches applied to medical imaging datasets for liver tumor analysis. From convolutional neural networks (CNNs) to more sophisticated architectures like UNet and attention mechanisms, this paper provides an in-depth analysis of the methodologies employed for tumor localization and boundary delineation. Additionally, it explores the integration of multimodal imaging data and transfer learning strategies to enhance model generalization across different datasets. Challenges, such as limited annotated data and interpretability issues, are discussed alongside potential solutions. The survey also highlights benchmark datasets and performance metrics commonly used in the field. Through this comprehensive examination, the novelty of this paper aims to offer insights into the current state-of-the-art, identify research gaps, and provide a roadmap for future developments in the domain of liver tumor detection and segmentation using deep learning models.