Midline shift (MLS) is a serious, potentially life-threatening displacement of brain structures. As in other medical fields, there is a need for computer-aided diagnostic systems to support rapid and accurate diagnosis in the detection of MLS. In this study, a comprehensive literature review was conducted by systematically searching academic papers published after 2010 in three databases: IEEE Xplore, PubMed, and Google Scholar. Only original research articles published in English, focusing on MLS detection using computed tomography (CT) images, were included in the review. Existing studies were reviewed under image processing (symmetry-based and landmark-based), machine learning (ML), and deep learning (DL) approaches. It was found that symmetry-based methods offer fast, basic evaluation; landmark-based methods are reliable with accurate landmark identification but can be affected by pathology. ML methods can perform well with large, well-processed datasets but may struggle with certain pathologies and anatomical variations, while DL methods offer high accuracy but require extensive data and may lose performance with rare cases. Computation time, asymmetry, pathological changes, image quality, artifacts, ground truth variability, data imbalance, and the limited number of cases have been identified as the main challenges in the detection of MLS. Solutions to overcome these challenges were also discussed in detail. This review provides a comprehensive overview of current methods and challenges in CT-based MLS detection to inform future research and clinical practice.