In the treatment of dermatological diseases, precise segmentation of lesion areas assists dermatologists in disease diagnosis and therapy. However, medical images often have low contrast or poor image quality due to blurring and noise, making it difficult for specialists to diagnose with just conventional visual observations. The problem of skin lesion segmentation using deep learning has been of interest and development recently, and has gained significant advancement with the progress of deep learning. In this study, we introduce a new model for skin lesion segmentation, named MambaFusionSeg, developed based on the fusion of Mamba, convolutional neural networks (CNNs), Priority Attention, and Depthwise Convolution into a unified framework. We propose two key components: Channel-Partitioned Mamba Block (CPM Block) and Spatially Adaptive Attention Block (SAAB) which improve spatial and channel-wise feature representation to enhance feature extraction and mitigate noise. We evaluate the performance of our MambaFusionSeg model on two popular skin lesion datasets, ISIC2018 and PH2, achieving state-of-the-art performance. Experimental results demonstrate the effectiveness of our approach in accurately delineating lesion boundaries while maintaining computational efficiency.