Aug 2026· Italian National Conference on Sensors· Vol 26· 0 citations· 42 references
Medicine
TL;DR
MGA-UNet, a frequency-aware multi-scale encoder–decoder segmentation framework that integrates wavelet-based frequency decomposition with Mamba-based long-range dependency modelling, demonstrates that frequency-domain decomposition and state-space modelling can complement each other for accurate medical image segmentation.
Abstract
Medical image segmentation is a critical task in computer-assisted diagnosis, but accurate delineation remains challenging in sensor-captured dermoscopic and endoscopic images because of low contrast, blurred boundaries, acquisition artifacts, and large appearance variations. Although CNN-based methods are effective in capturing local texture and boundary cues, they often struggle to explicitly model long-range dependencies and global structural relationships. Transformer-based architectures can capture global context, but their self-attention mechanism may become computationally costly when processing high-resolution feature maps. To address these challenges, we propose MGA-UNet, a frequency-aware multi-scale encoder–decoder segmentation framework that integrates wavelet-based frequency decomposition with Mamba-based long-range dependency modelling. Specifically, the Wavelet-Mamba feature extraction backbone (WMB) decomposes features into low- and high-frequency components to enhance boundary-aware representation, the Gated Multi-scale Aggregation Module (GMAM) aggregates parallel multi-scale encoder features and applies a content-dependent gate to the fused response, and the Adaptive Sparse Attention Module (ASAM) refines bottleneck representations with sparse attention for global semantic modelling. Across three independent runs with random seeds 42, 123, and 2026, MGA-UNet achieves mean Dice Similarity Coefficients of 88.92±0.04%, 88.01±0.07%, and 85.91±0.04% on ISIC2018, ISIC2017, and Kvasir-SEG, respectively. These results demonstrate competitive segmentation performance among the compared representative CNN-based, Transformer-based, and Mamba-based methods, including the recent H-VMUNet baseline. These results indicate that frequency-domain decomposition and state-space modelling can complement each other for accurate medical image segmentation, particularly in images with ambiguous boundaries and complex background interference.
Results indicate the WVM-UNet architecture effectively captures discriminative features for precise medical image segmentation, and demonstrates the competitive performance of the method on multiple public datasets.
Yulong Yang, Wenchao Gao, Zheng-Guo Wu et al.· Journal of Imaging· 0 citations
DIG-MambaNet achieves consistent and competitive performance across diverse target structures and imaging conditions, with improved boundary delineation and favorable overlap-based accuracy compared with representative CNN-, Transformer-, and Mamba-based methods.
Yongkang Zhu, Tianyu Yu, Hongmei Li et al.· Journal of Imaging· 0 citations
GLNet adopts a dual-branch encoder that combines a CNN-based Local Detail Perception Branch with a Mamba-based Global Context Modeling Branch, enabling the joint extraction of fine-grained local features and long-range semantic representations.
Accurate boundary delineation remains a persistent challenge in dermoscopic image segmentation because of blurred lesion margins, heterogeneous textures, and complex background artifacts. From a signal-processing perspective, lesion boundaries represent high-frequency components that are highly susceptible to aliasing, noise amplification, and information loss. Consequently, repeated downsampling and feature transformations in conventional convolutional architectures often lead to severely degraded boundary representations. To address these limitations, we propose EA-LiteUNet, an edge-adaptive and computationally efficient U-Net variant specifically designed for boundary-sensitive medical image segmentation. The architecture integrates three core mechanisms: (1) boundary-aware representation learning to suppress aliasing and preserve high-frequency structural details; (2) attention-guided feature modulation to selectively enhance boundary-relevant responses across multi-scale features; and (3) a resource-adaptive inference strategy to dynamically balance segmentation accuracy and computational efficiency. Extensive evaluations across three public dermoscopic datasets demonstrate that EA-LiteUNet consistently achieves superior boundary precision. Specifically, on the ISIC 2018 dataset, the method significantly reduces the 95% Hausdorff Distance (HD95) to 12.89 pixels while maintaining a robust Dice score of 92.08%. Notably, this strong performance is achieved with an ultralightweight configuration of merely 0.29M parameters and 1.17 GFLOPs. Ablation studies further validate the complementary effects of these components, confirming their contribution to enhanced boundary fidelity and stable optimization.
Jiangtao Wang, N. Ruhaiyem, Panpan Fu et al.· 0 citations
Automatic skin lesion segmentation is a pivotal problem in the medical domain and an indispensable component in the computer-aided diagnosis program. Most convolutional neural network-based segmentation algorithms have demonstrated promising performance due to their ability to encode detail and semantic features efficiently. However, they fail to capture the long-range contextual information at the global level. Therefore, researchers employ Transformer architecture to address this issue. Unfortunately, these methods fail to learn sufficient pixel information at the local level. Motivated by this, some researchers attempt to design a hybrid architecture based on CNN and Transformer. However, the large number of parameters and high computational cost make them challenging to train and use. To alleviate these problems, we propose an effective Lightweight Multi-scale Deformation Aggregation U-Net (LMDAU-Net), which consists of a Lightweight Local-global Learning Module (LLM) and an Adaptive Interactive Fusion Module (AIF). Specifically, we utilize the two branches of the proposed LLM to efficiently learn local fine-grained and global coarse-grained features that assist the model in capturing the complementary feature representations. Moreover, we employ the AIF to selectively learn semantic and detail features at different scales, which can dynamically explore variable feature cues. Extensive experiments on four skin benchmarks, including ISIC2016, ISIC 2017, ISIC2018, and PH2, demonstrate that LMDAU-Net achieves state-of-the-art performance in both qualitative and quantitative aspects. We have released our code on https://github.com/Lm0611/LMDAU-Net.
Jun-Han Hu, Ming Liu, Jing Yang et al.· IEEE journal of biomedical a...· 0 citations
