<p>The core task of skin lesion segmentation involves precise localization of lesion regions within images, as segmentation accuracy directly influences clinical diagnosis and treatment outcomes. Due to the low contrast between skin lesions and normal tissues, the non-uniform color of lesion regions, as well as blurred and irregular boundaries, makes accurate segmentation difficult. Motivated by these challenges, we present a Boundary-Aware Mamba Network for skin lesion segmentation to cope with the challenges of complex scenarios in segmentation tasks. This method innovatively designs a frequency-enhanced visual state space block as the basic block of the network, which not only captures extensive contextual information but also strengthens the high-frequency texture information of lesion boundaries. We design a CLAHE edge enhancement module to directly perform local contrast enhancement on skin lesion images. Meanwhile, we design an attention-guided multi-feature fusion module that combines attention mechanisms with residual networks to optimize multi-feature fusion. We have conducted rigorous experimental validation on four open-source datasets (ISIC2016, ISIC2017, ISIC2018 and PH2) and performed comparative evaluations with advanced existing methods. The results confirm that our method can produce more precise and detailed segmentation results in skin lesion segmentation tasks.</p>