The field of medical image segmentation is rapidly advancing with the development of innovative models and techniques. Recent research has focused on improving the accuracy and efficiency of segmentation models, particularly in resource-limited settings. One notable trend is the use of lightweight and promptable models, such as those utilizing U-Net-style architectures and attention mechanisms, which have shown promising results in segmenting 3D medical images. Another area of research is the development of unified modality-relax segmentation networks, which can effectively handle missing or corrupted modalities and improve segmentation performance. Additionally, there has been a growing interest in exploring the potential of hyperspectral imaging and vision foundation models for robust robotic perception and semantic segmentation. Noteworthy papers in this area include ENSAM, which achieved state-of-the-art performance in the CVPR 2025 Foundation Models for Interactive 3D Biomedical Image Segmentation Challenge, and UniMRSeg, which proposed a unified modality-relax segmentation network through hierarchical self-supervised compensation. MK-UNet and SSCM also demonstrated significant improvements in medical image segmentation and multi-contrast MRI super-resolution, respectively. HiPerformer introduced a novel modular hierarchical fusion strategy for global-local segmentation, and the Hyperspectral Adapter achieved state-of-the-art semantic segmentation performance using hyperspectral inputs.