<p>In recent years, diffusion models have shown great potential in generating high-quality images for medical image analysis. However, existing methods [such as DDPM (Ho et al., Adv Neural Inf Process Syst 33:6840–6851, 2020) and Stable Diffusion (Lin et al., arXiv preprint <a href="https://arxiv.org/abs/2406.18361">arXiv:2406.18361</a>, 2024)] still face several challenges, including the loss of fine details and edge information during the denoising process, incomplete reconstruction of complex anatomical structures (such as vascular bifurcations and tumor infiltration), and insufficient modeling of multi-scale long-range dependencies. To address these issues, this paper proposes FRF-SEDNet (Feature ReAssembly and Refined Differential Edge-aware Network based on Stable Diffusion), which improves the performance of medical image segmentation through three innovative modules: 1. The xLSTM-UNet architecture is introduced to address the issue of long-range dependency disruption caused by iterative denoising during the diffusion process (such as the continuity of anatomy between consecutive slices), thereby improving the coherence of tumor boundaries. Experimental results show that the boundary coherence improves by 23% (Dice<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\uparrow \)</EquationSource> </InlineEquation>). 2. A cascade module composed of EFDM and EGCM is designed to explicitly enhance edge gradient responses and utilize neighborhood topology relationships to correct confusing regions (such as pancreatic adhesion with the intestinal tract). 3. A Feature-Aware Reassembly Module (FERF) is employed to optimize the transformation of latent representations into high-resolution images, suppressing common artifacts and blurry artifacts in diffusion generation. On the ISIC2016, ISIC2017, and ISIC2018 datasets, the Dice coefficients of FRF-SEDNet are 93.84%, 90.32%, and 90.82%, respectively. On the TN3K dataset (used for thyroid ultrasound image segmentation) and the BrainMRI dataset (used for brain tumor segmentation), the Dice coefficients reach 87.5% and 88.2%, surpassing classic models such as UNet and SSFormer, as well as outperforming DermoSegDiff and MedSegDiff in the same category of diffusion models. Additionally, the inference speed is improved by a factor of 3 (with single-image inference requiring only 1.1 s). Furthermore, the model was trained on the BrainMRI dataset and generalized on the Kvasir dataset, achieving a Dice score of 90.96%, demonstrating strong generalization performance. FRF-SEDNet demonstrates excellent performance in medical image segmentation while maintaining high inference efficiency. It is poised to become a powerful and effective tool for medical image segmentation, providing significant support to physicians in diagnosis.</p>

FRF-SEDNet: feature reassembly and refined differential edge-aware network based on stable diffusion

期刊论文
发表日期：

2026年2月2日

查看全文

Chengsen Xu,
Peng Duan,
Jinjiang Li

摘要

In recent years, diffusion models have shown great potential in generating high-quality images for medical image analysis. However, existing methods [such as DDPM (Ho et al., Adv Neural Inf Process Syst 33:6840–6851, 2020) and Stable Diffusion (Lin et al., arXiv preprint arXiv:2406.18361, 2024)] still face several challenges, including the loss of fine details and edge information during the denoising process, incomplete reconstruction of complex anatomical structures (such as vascular bifurcations and tumor infiltration), and insufficient modeling of multi-scale long-range dependencies. To address these issues, this paper proposes FRF-SEDNet (Feature ReAssembly and Refined Differential Edge-aware Network based on Stable Diffusion), which improves the performance of medical image segmentation through three innovative modules: 1. The xLSTM-UNet architecture is introduced to address the issue of long-range dependency disruption caused by iterative denoising during the diffusion process (such as the continuity of anatomy between consecutive slices), thereby improving the coherence of tumor boundaries. Experimental results show that the boundary coherence improves by 23% (Dice \(\uparrow \) ). 2. A cascade module composed of EFDM and EGCM is designed to explicitly enhance edge gradient responses and utilize neighborhood topology relationships to correct confusing regions (such as pancreatic adhesion with the intestinal tract). 3. A Feature-Aware Reassembly Module (FERF) is employed to optimize the transformation of latent representations into high-resolution images, suppressing common artifacts and blurry artifacts in diffusion generation. On the ISIC2016, ISIC2017, and ISIC2018 datasets, the Dice coefficients of FRF-SEDNet are 93.84%, 90.32%, and 90.82%, respectively. On the TN3K dataset (used for thyroid ultrasound image segmentation) and the BrainMRI dataset (used for brain tumor segmentation), the Dice coefficients reach 87.5% and 88.2%, surpassing classic models such as UNet and SSFormer, as well as outperforming DermoSegDiff and MedSegDiff in the same category of diffusion models. Additionally, the inference speed is improved by a factor of 3 (with single-image inference requiring only 1.1 s). Furthermore, the model was trained on the BrainMRI dataset and generalized on the Kvasir dataset, achieving a Dice score of 90.96%, demonstrating strong generalization performance. FRF-SEDNet demonstrates excellent performance in medical image segmentation while maintaining high inference efficiency. It is poised to become a powerful and effective tool for medical image segmentation, providing significant support to physicians in diagnosis.