Foundation Models for Medical Image Segmentation: A Literature Review

Medical image segmentation is an important component of medical image analysis, allowing precise delineation of regions of interest for accurate diagnosis and treatment planning. Deep learning models have traditionally dominated this field; however, their reliance on large task-specific datasets and limited generalization capabilities present significant challenges. The emergence of foundation models (FMs), particularly the Segment Anything Model (SAM), has introduced a new paradigm by enabling vision FMs to perform diverse segmentation tasks without the need for re-training. This study presents a system-atic literature review of FMs for medical image segmentation, synthesizing 27 papers published between 2023 and 2024. The review examines three key aspects: the development of novel FMs designed for medical image segmentation, adaptations of SAM for medical imaging applications, and primary challenges asso-ciated with implementing FMs in this domain. By consolidating recent advances and limitations, this study provides an updated perspective on the role of FMs in medical-image segmentation.