LightM-UNet is a lightweight medical image segmentation model that integrates Mamba, a State Space Model, into the UNet architecture to address computational limitations in real-world medical applications. Traditional UNet and its variants, especially those based on Transformers, suffer from high parameter counts and computational costs, making them unsuitable for mobile healthcare. LightM-UNet replaces CNN and Transformer components with Mamba, achieving state-of-the-art performance with significantly reduced parameters and computational costs. It uses a Residual Vision Mamba Layer (RVM Layer) to extract deep semantic features and model long-range spatial dependencies with linear computational complexity. LightM-UNet outperforms existing models like nnU-Net and U-Mamba, achieving superior segmentation performance while reducing parameters by 116x and computational costs by 21x. The model is validated on two real-world datasets, demonstrating its effectiveness in 2D and 3D medical image segmentation. LightM-UNet's lightweight design makes it suitable for mobile healthcare applications, offering a promising solution for efficient medical image segmentation. The model's architecture includes Encoder Blocks, a Bottleneck Block, and Decoder Blocks, with extensive experiments showing its superiority in performance and efficiency. The study highlights the potential of Mamba in enabling lightweight and efficient medical image segmentation models.LightM-UNet is a lightweight medical image segmentation model that integrates Mamba, a State Space Model, into the UNet architecture to address computational limitations in real-world medical applications. Traditional UNet and its variants, especially those based on Transformers, suffer from high parameter counts and computational costs, making them unsuitable for mobile healthcare. LightM-UNet replaces CNN and Transformer components with Mamba, achieving state-of-the-art performance with significantly reduced parameters and computational costs. It uses a Residual Vision Mamba Layer (RVM Layer) to extract deep semantic features and model long-range spatial dependencies with linear computational complexity. LightM-UNet outperforms existing models like nnU-Net and U-Mamba, achieving superior segmentation performance while reducing parameters by 116x and computational costs by 21x. The model is validated on two real-world datasets, demonstrating its effectiveness in 2D and 3D medical image segmentation. LightM-UNet's lightweight design makes it suitable for mobile healthcare applications, offering a promising solution for efficient medical image segmentation. The model's architecture includes Encoder Blocks, a Bottleneck Block, and Decoder Blocks, with extensive experiments showing its superiority in performance and efficiency. The study highlights the potential of Mamba in enabling lightweight and efficient medical image segmentation models.