Swin-UMamba is a Mamba-based model designed for medical image segmentation, leveraging ImageNet-based pretraining to enhance performance. The model integrates a Mamba-based encoder pretrained on ImageNet to extract multi-scale features and a decoder for segmentation tasks. It also introduces a variant, Swin-UMamba†, with a Mamba-based decoder that reduces parameters and computational load while maintaining competitive performance. Experimental results show that Swin-UMamba outperforms CNNs, ViTs, and other Mamba-based models on datasets like AbdomenMRI, Endoscopy, and Microscopy, with an average improvement of 2.72% over its closest counterpart U-Mamba_Enc. ImageNet-based pretraining significantly improves segmentation accuracy, convergence stability, and data efficiency. The model's performance is validated across three medical image segmentation datasets, demonstrating its effectiveness in handling long-range dependencies and complex medical images. The study highlights the importance of pretraining in enhancing the performance of Mamba-based models for medical imaging tasks.Swin-UMamba is a Mamba-based model designed for medical image segmentation, leveraging ImageNet-based pretraining to enhance performance. The model integrates a Mamba-based encoder pretrained on ImageNet to extract multi-scale features and a decoder for segmentation tasks. It also introduces a variant, Swin-UMamba†, with a Mamba-based decoder that reduces parameters and computational load while maintaining competitive performance. Experimental results show that Swin-UMamba outperforms CNNs, ViTs, and other Mamba-based models on datasets like AbdomenMRI, Endoscopy, and Microscopy, with an average improvement of 2.72% over its closest counterpart U-Mamba_Enc. ImageNet-based pretraining significantly improves segmentation accuracy, convergence stability, and data efficiency. The model's performance is validated across three medical image segmentation datasets, demonstrating its effectiveness in handling long-range dependencies and complex medical images. The study highlights the importance of pretraining in enhancing the performance of Mamba-based models for medical imaging tasks.