发表机构
School of Computer Science and Technology, Zhejiang Gongshang University; Tongxiang Hospital of Traditional Chinese Medicine; Prince Sattam Bin Abdulaziz University; Qena University; Tampere University; Universidade Federal Fluminense; Federal University of Ceara; Firat University; Systems Research Institute, Polish Academy of Sciences; AGH University of Krakow（浙江工商大学计算机科学与技术学院; 桐乡市中医院; 萨塔姆·本·阿卜杜勒阿齐兹王子大学; 奎纳大学; 坦佩雷大学; 弗卢米嫩塞联邦大学; 塞阿拉联邦大学; 菲拉特大学; 波兰科学院系统研究所; 克拉科夫AGH大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对口腔扫描图像分割中现有视觉SSMs破坏空间连续性、成像干扰影响边界定位的问题，提出融合动态扫描与频域增强的FU-Mamba，在牙科数据集上使mIoU提升1.1%。
AI 中文摘要
口腔扫描图像分割是数字牙科中计算机辅助诊断和治疗规划的关键环节。然而，现有的视觉状态空间模型（SSMs）通常依赖人工设计的扫描顺序将图像块展平为序列，这会破坏语义空间连续性，阻碍从关键前景区域提取连贯特征。此外，数据采集过程中的光照不一致、反射表面、噪声等因素会通过削弱高频细节、增强低频分量来扰乱频率分布，进而阻碍边界的精确定位。针对这些挑战，我们提出了FU-Mamba，一种在SSM架构中融合动态扫描与频域增强的创新框架。具体而言，动态Mamba模块（DMB）通过可训练的偏移预测网络自适应学习采样偏移，并执行灵活的双线性插值，实现保留空间连贯性的内容感知扫描。此外，频域增强模块通过小波引导分解和频谱池化平衡频谱分量，提升在不利成像条件下的鲁棒性。实验结果表明，在牙科分割数据集上评估时，FU-Mamba的平均交并比（mIoU）指标提升了1.1%，分割精度得到显著增强。项目页面：this https URL
英文摘要
Oralscan image segmentation is essential for computer-aided diagnosis and treatment planning in digital dentistry. However, existing visual state space models (SSMs) often rely on manually designed scanning orders to flatten image patches into sequences, which disrupts the semantic spatial continuity and hinders coherent feature extraction from key foreground regions. Moreover, elements such as inconsistent lighting, reflective surfaces, and noise during data acquisition disrupt the frequency distribution by diminishing high-frequency details while enhancing low-frequency components, consequently hindering the accurate localization of boundaries. In response to these challenges, we introduce FU-Mamba, an innovative framework that incorporates dynamic scanning and frequency domain enhancement within the SSM architecture. Specifically, the Dynamic Mamba Block (DMB) adaptively learns sampling offsets via a trainable offset prediction network and performs flexible bilinear interpolation, enabling content-aware scanning that preserves spatial coherence. Furthermore, a frequency domain enhancement block balances spectral components through wavelet-guided decomposition and spectrum pooling, improving robustness under adverse imaging conditions. Experimental findings indicate that FU-Mamba attains a notable enhancement in segmentation accuracy, evidenced by a 1.1% increase in the mean intersection over union (mIoU) metric when evaluated on the dental segmentation dataset. Project page: https://byte2bite.github.io/FU-Mamba/
CommentsAccepted by Neurocomputing
Journal refNeurocomputing, Volume 701, 2026, 134618
DOI：10.1016/j.neucom.2026.134618
