摘要
针对磁共振图像中存在病变器官组织复杂多变和提取不完全的问题,提出一种基于Mamba和卷积神经网络的医学图像分割方法.首先,使用卷积神经网络模块以构建多级局部特征之间的相关性;其次,使用Mamba构建主干网络,增强远程建模能力,有效捕获图像中的全局上下文信息;最后,提出基于通道注意力引导的自适应特征融合模块,在抑制背景噪声干扰的同时实现相邻层次特征之间的有效融合.在Synapse数据集上进行实验验证,该算法mIoU和mDice分别达到72.64%、83.35%.优于与之对比的其他分割方法,能够实现更精准的医学图像分割,从而辅助医生准确判断病变情况.
Aiming at the complex and variable appearance of pathological tissues in magnetic resonance imaging,along with incomplete feature extraction,this study proposes a novel medical image segmentation framework that integrates Mamba architecture with convolutional neural networks.First,a CNN-based module is designed to capture multi-scale local feature dependencies.Then,a Mamba-based backbone network is introduced to enhance long-range modeling capability and effectively capture global contextual information in medical images.Finally,an adaptive feature fusion module guided by channel attention is proposed to suppress interference steming from background noise while enabling effective integration of adjacent hierarchical features.Extensive experiments on the synapse multi-organ segmentation dataset demonstrate that the proposed method achieves a mIoU of 72.64%and a mDice of 83.35%,and outperforms several state-of-the-art segmentation approaches.The results indicate that our method enables more precise medical image segmentation,thereby providing a reliable assistance for clinicians in accurate pathological assessment.
作者
马瀚词
陈辉
MA Hanci;CHEN Hui(School of Computer Science and Engineering,Anhui University of Science and Technology,Huainan Anhui 232001)
出处
《宁夏师范大学学报》
2026年第1期91-102,共12页
Journal of Ningxia Normal University
基金
安徽省重点教学研究项目(2020jyxm0458).
