<p>Accurate segmentation of skin lesions is crucial for early melanoma detection and improved prognosis. However, this task remains challenging due to issues such as hair interference, low contrast, and blurred boundaries in dermoscopic images. To address these issues, we propose IMANet, an Iterative Multi-scale Attention Network that introduces significant advancements in lesion segmentation. IMANet leverages three key innovations: the Efficient Residual Attention (ERA) module, which enhances feature representation through channel-wise attention and efficient convolution with residual learning; the Dual Path Masked Fusion (DPMF) module, which captures multi-scale features using diverse kernel sizes and dynamically adjusts their importance via a masking mechanism; and the Gated Feature Fusion (GFF) module, which integrates contextual information effectively by adaptively fusing features across scales. Experiments conducted on the public ISIC2017 and ISIC2018 dataset demonstrate the effectiveness of IMANet, achieving superior performance compared to state-of-the-art methods. Our approach not only improves segmentation accuracy but also enhances the visualization of lesion boundaries, which is crucial for clinical applications. The implementation can be found at: <a href="https://github.com/bi-H-m/IMANet">https://github.com/bi-H-m/IMANet</a></p>