脑肿瘤属于神经系统中有较高致死率的疾病,其精准诊断和治疗的水平直接对患者的预后情况及生存质量产生影响。卷积神经网络在图像特征提取与分割中展现出高效的特征学习能力与分割精确性,为脑肿瘤的快速分割提供了高效的解决方案。针对脑肿瘤MRI图像分割过程中出现的边界模糊、肿瘤和正常组织对比度低、背景噪声等问题,在传统U-Net架构中融合残差块与注意力机制,构建基于残差注意力的改进U-Net模型。实验采用公开数据集训练模型,通过Sigmoid激活函数与二元交叉熵损失函数联合优化,缓解肿瘤与正常组织的类别不均衡问题。模型经多轮迭代优化后在测试集MRI图像上完成分割验证,结果显示模型分割结果与真实标签相吻合,且在量化指标上均取得良好表现。消融实验结果表明,残差注意力U-Net与传统U-Net、残差U-Net、注意力U-Net的分割性能相比,在Dice、IoU、Recall等指标上更具优势,充分验证了该模型在脑肿瘤MRI图像分割任务中的有效性和实用价值。
Abstract
Brain tumors are among the most lethal diseases in the nervous system, and the accuracy of their diagnosis and treatment directly impacts patient prognosis and quality of life. Convolutional Neural Networks demonstrate efficient feature learning capabilities and high segmentation precision in image feature extraction and segmentation, providing an efficient solution for fast brain tumor segmentation. To address challenges in brain tumor MRI image segmentation, such as blurred boundaries, low contrast between tumors and normal tissue, and background noise, we integrate residual blocks and attention mechanisms into the traditional U-Net architecture to develop an enhanced U-Net model based on residual attention. The model is trained using a public dataset, optimized jointly with the Sigmoid activation function and binary cross entropy loss function to alleviate class imbalance between tumor and normal tissue. After multiple rounds of iterative optimization, the model achieves segmentation validation on test MRI images, with results closely matching ground truth labels and excellent performance across quantitative metrics. Ablation experiments demonstrate that the residual attention U-Net outperforms traditional U-Net, residual U-Net, attention U-Net in terms of Dice, IoU, and Recall metrics, confirming the model's effectiveness and practical value in brain tumor MRI image segmentation tasks.
关键词
脑肿瘤 /
医学图像分割 /
U-Net模型 /
残差块 /
注意力机制
Key words
brain tumor /
medical image segmentation /
U-Net model /
residual block /
attention mechanism
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Solanki S,Singh U P,Chouhan S S,et al.Brain tumor detection and classification using intelligence techniques:an overview[J].IEEE Access,2023,11:12870-12886.
[2] Ranjbarzadeh R,Caputo A,Tirkolaee E B,et al.Brain tumor segmentation of MRI images:A comprehensive review on the application of artificial intelligence tools[J].Computers in biology and medicine,2023,152:106405.
[3] Abdou M A.Literature review:Efficient deep neural networks techniques for medical image analysis[J].Neural Computing and Applications,2022,34(8):5791-5812.
[4] Chen S,Qiu C,Yang W,et al.Combining edge guidance and feature pyramid for medical image segmentation[J].Biomedical signal processing and control,2022,78:103960.
[5] Jardim S,António J,Mora C.Image thresholding approaches for medical image segmentation-short literature review[J].Procedia Computer Science,2023,219:1485-1492.
[6] Lin Q,Chen X,Chen C,et al.Boundary-wise loss for medical image segmentation based on fuzzy rough sets[J].Information Sciences,2024,661:120183.
[7] Karimi D,Warfield S K,Gholipour A.Transfer learning in medical image segmentation:New insights from analysis of the dynamics of model parameters and learned representations[J].Artificial intelligence in medicine,2021,116:102078.
[8] Chen J,Wan Z,Zhang J,et al.Medical image segmentation and reconstruction of prostate tumor based on 3D AlexNet[J].Computer methods and programs in biomedicine,2021,200:105878.
[9] Sistaninejhad B,Rasi H,Nayeri P.A review paper about deep learning for medical image analysis[J].Computational and Mathematical Methods in Medicine,2023,2023(1):7091301.
[10] Iqbal S,Naqvi S S,Khan H A,et al.G-Net Light:A lightweight modified Google Net for retinal vessel segmentation[C]//Photonics.MDPI,2022,9(12):923.
[11] Azad R,Aghdam E K,Rauland A,et al.Medical image segmentation review:The success of u-net[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2024.
[12] 刘书睿. 基于U-Net的医学图像分割模型优化综述[J].电子技术,2025,54(4):152-154.
[13] 付豪,张振利,陈源.改进UNet++模型的脑肿瘤图像分割算法[J].河北大学学报(自然科学版),2025,45(4):398-407.
[14] 卢力玮,汪洋,柳杨,等.基于多尺度通道融合注意力的皮肤癌U-Net分割模型[J/OL].信息与控制,1-12.DOI:10.13976/j.cnki.xk.2024.3231.
[15] Aghalari M,Aghagolzadeh A,Ezoji M.Brain tumor image segmentation via asymmetric/symmetric UNet based on two-pathway-residual blocks[J].Biomedical signal processing and control,2021,69:102841.
[16] Cai Y,Wang Y.Ma-unet:An improved version of unet based on multi-scale and attention mechanism for medical image segmentation[C]//Third international conference on electronics and communication; network and computer technology (ECNCT 2021).SPIE,2022,12167:205-211.
基金
南京医科大学康达学院科研发展基金项目,项目编号：KD2023KYJJ026; 南京医科大学康达学院第二期品牌专业建设资助项目,项目编号：JX206000302
