Conceptual

Hybrid CNN-Mamba Architectures for Medical Image Segmentation

Deep-learning segmentation networks that combine convolutional neural network backbones with state-space (Mamba) modules, often alongside attention mechanisms, to segment structures such as tumors in medical images. The goal is to capture both local detail (via convolutions) and long-range context (via linear-time state-space sequence modeling) while keeping parameters and computation low, addressing the heterogeneous morphology and indistinct boundaries typical of medical imaging.