MRFA-Net：一种基于多尺度残差特征挖掘的肺癌CT图像分割网络

MRFA-Net: a lung cancer CT image segmentation network based on multi-scale residual feature mining

摘要

医学影像中的肺部肿瘤分割对于辅助诊断和治疗规划具有重要意义。然而，由于肿瘤在形态、边界和大小上的高度异质性，精准分割仍然具有较大挑战。提出一种基于残差特征挖掘与深层特征融合的深度分割网络MRFA-Net，旨在提升肺肿瘤在CT图像中的自动分割性能。提出多尺度残差特征提取模块和深层特征聚合模块，通过在多尺度上对残差特征进行提取以及深层特征融合，更有效地挖掘出具有代表性的肿瘤特征，并且在跳跃连接中引入反向注意力机制以提升模型对关键区域的感知能力。在MSD公开数据集和私有肺癌CT数据集上进行的实验结果表明MRFA-Net具备优异性能，Dice相似系数分别达到74.15%和74.77%。该方法为肺部肿瘤分割提供有效的解决方案，具备良好的临床应用潜力。

Abstract

Lung tumor segmentation from medical images is crucial for supporting diagnosis and treatment planning. However, tumors exhibit substantial heterogeneity in shape, contour and size, which makes precise segmentation still challenging. Therefore, this paper proposes a deep segmentation network named MRFA-Net (multi-scale residual feature aggregation network). Built upon residual feature mining and deep feature fusion, the network aims to improve the automatic segmentation performance of lung tumors on CT images. Specifically, two key modules are designed: the multi-scale residual feature extraction module and the deep pyramid aggregation module. By extracting residual features at multiple scales and fusing deep features, the network can capture representative tumor characteristics more effectively. Additionally, a reverse attention mechanism is integrated into skip connections to enhance the model's perception of critical regions. Experimental results on the public MSD dataset and a private lung cancer CT dataset demonstrate that MRFA-Net achieves excellent performance, with a Dice similarity coefficient of 74.15% and 74.77%, respectively. The proposed method provides an effective solution for lung tumor segmentation and holds promising prospects for clinical practice.

关键词

肺部肿瘤分割 / 残差特征提取 / 深层特征聚合 / 反向注意力机制

Key words

lung tumor segmentation / residual feature extraction / deep pyramid aggregation / reverse attention mechanism

参考文献

[1] 岳文怡, 盛复庚 . 基于磁共振成像的乳腺癌影像基因组学的研究现状和进展［J］. 临床放射学杂志, 2024, 43(6): 991-994.

[2] Yue WY, Sheng FG . Research status and progress of breast cancer radiogenomics based on magnetic resonance imaging［J］. Journal of Clinical Radiology, 2024, 43(6): 991-994.

[3] 周力凯, 江雨洋, 冯亚春, 等 . 基于多尺度区域与类不确定性理论的局部阈值分割方法［J］. 计算机应用, 2020, 40(增刊2): 66-72.

[4] Zhou LK, Jiang YY, Feng YC, et al. Local image thresholding method based on multiscale region and class uncertainty theory［J］. Journal of Computer Applications, 2020, 40(S2): 66-72.

[5] 邵浩辰, 张伟, 马宇张 . 基于ResUNet和Transformer的CT肝脏肿瘤图像分割［J］. 中国医学物理学杂志, 2025, 42(11): 1455-1461.

[6] Shao HC, Zhang W, Ma YZ . CT liver tumor image segmentation based on ResUNet and Transformer［J］. Chinese Journal of Medical Physics, 2025, 42(11): 1455-1461.

[7] Ronneberger O, Fischer P, Brox T . U—Net: convolutional networks for biomedical image segmentation［C］// Medical Image Computing and Computer—Assisted Intervention—MICCAI 2015. Cham: Springer International Publishing, 2015: 234-241.

[8] Xiao X, Lian S, Luo ZM, et al. Weighted Res—UNet for high—quality retina vessel segmentation［C］// 2018 9th International Conference on Information Technology in Medicine and Education (ITME). Piscataway, NJ, USA: IEEE, 2018: 327-331.

[9] Dong RS, Pan XQ, Li FY . DenseU—Net—based semantic segmentation of small objects in urban remote sensing images［J］. IEEE Access, 2019, 7: 65347-65356.

[10] Ding QY, Pan Y, Liu JX, et al. AMEF—Net: towards an attention and multi—level enhancement fusion for medical image classification in Parkinson's aided diagnosis［J］. IET Computer Vision, 2024, 19(1): e12324.

[11] Fan DP, Ji GP, Zhou T, et al. PraNet: parallel reverse attention network for polyp segmentation［C］// Medical Image Computing and Computer Assisted Intervention—MICCAI 2020. Cham: Springer International Publishing, 2020: 263-273.

[12] Isensee F, Jaeger PF, Kohl SAA, et al. nnU—Net: a self—configuring method for deep learning—based biomedical image segmentation［J］. Nat Methods, 2021, 18(2): 203-211.

[13] Shi LX, He WY, Wang GD . A dual—path fusion network with edge feature enhancement for medical image segmentation［J］. Mathematics, 2026, 14(1): 55.

[14] Hatamizadeh A, Nath V, Tang YC, et al. Swin UNETR: Swin transformers for semantic segmentation of brain tumors in MRI images［C］// Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. Cham: Springer International Publishing, 2022: 272-284.

[15] Zhang YD, Liu HY, Hu Q . TransFuse: fusing transformers and CNNs for medical image segmentation［C］// Medical Image Computing and Computer Assisted Intervention—MICCAI 2021. Cham: Springer International Publishing, 2021: 14-24.

[16] Yuan L, Chen YP, Wang T, et al. Tokens—to—token ViT: training vision transformers from scratch on ImageNet［C］// 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway, NJ, USA: IEEE, 2021: 538-547.

[17] Yang L, Zhang JW, Wang T, et al. Multi—scale camouflaged feature mining and fusion network for liver tumor segmentation［J］. Eng Appl Artif Intell, 2025, 148: 110398.

[18] Su J, Liu Z, Zhang J, et al. DV—Net: accurate liver vessel segmentation via dense connection model with D—BCE loss function［J］. Knowl Based Syst, 2021, 232: 107471.

[19] Shit S, Paetzold JC, Sekuboyina A, et al. clDice—a novel topology—preserving loss function for tubular structure segmentation［C］// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2021: 16555-16564.

[20] 李翠锦, 瞿中 . 基于卷积神经网络的跨层融合边缘检测算法［J］. 计算机应用研究, 2021, 38(7): 2183-2187.

[21] Li CJ, Qu Z . Cross fusion edge detection algorithm based on CNN［J］. Application Research of Computers, 2021, 38(7): 2183-2187.

[22] Fu XH, Bi L, Kumar A, et al. Multimodal spatial attention module for targeting multimodal PET—CT lung tumor segmentation［J］. IEEE J Biomed Health Inform, 2021, 25(9): 3507-3516.

[23] Zhang YJ, Chen X . Lightweight semantic segmentation algorithm based on MobileNetV3 network［C］// 2020 International Conference on Intelligent Computing, Automation and Systems (ICICAS). Piscataway, NJ, USA: IEEE, 2020: 429-433.

[24] Qamar S, Fazil M, Ahmad P, et al. UNet with self—adaptive Mamba—like attention and causal—resonance learning for medical image segmentation［J］. Sci Rep, 2025, 16(1): 135.

[25] Wang B, Shi H, Zhao ZT, et al. PSCT—Net: a parallel symmetric CNN—transformer hybrid network for medical image segmentation［J］. Med Eng Phys, 2025, 146: 104442.

[26] Peng HX, Xue C, Shao YY, et al. Semantic segmentation of litchi branches using DeepLabV3+ model［J］. IEEE Access, 2020, 8: 164546-164555.

[27] 李柯, 刘文忠, 秦镜淘 . 改进MambaUNet网络对肝脏肿瘤CT图像的轻量化级联分割［J］. 中国医学物理学杂志, 2025, 42(8): 1068-1078.

[28] Li K, Liu WZ, Qin JT . Improved MambaUNet for lightweight cascaded segmentation in liver tumor CT image［J］. Chinese Journal of Medical Physics, 2025, 42(8): 1068-1078.

[29] Tagnamas J, Ramadan H, Yahyaouy A, et al. CS—Net: combined ConvNeXt—Swin—Unet for accurate medical image segmentation［J］. J Supercomput, 2026, 82(5): 276.

[30] Yang YW, Huang HG, Zhang GD, et al. DHR—Net: an ultra—lightweight U—Net based on efficient convolutional attention for medical image segmentation［J］. Displays, 2026, 93: 103425.

[31] Lei LH, Li W . Transformer—based multi—task model for lung tumor segmentation and classification in CT images［J］. J Radiat Res Appl Sci, 2025, 18(3): 101657.

[32] Yi C, Jiang SF, Xiong LL, et al. D—S—Net: an efficient dual—stage strategy for high—precision segmentation of gross tumor volumes in lung cancer CT images［J］. BMC Cancer, 2025, 25(1): 1387.

[33] Dwivedi P, Barage S, Singh R, et al. PET—based radiomic analysis in multicentre lung cancer study and impact of feature domain harmonization［J］. Phys Eng Sci Med, 2025, 48(4): 1841-1851.

[34] 付宏建, 白宏阳, 郭宏伟, 等 . 融合多注意力机制的光学遥感图像目标检测方法［J］. 光子学报, 2022, 51(12): 304-312.

[35] Fu HJ, Bai HY, Guo HW, et al. Object detection method of optical remote sensing image with multi—attention mechanism［J］. Acta Photonica Sinica, 2022, 51(12): 304-312.

基金资助

上海市科委科技创新行动计划(23010501700)

江西省卫健委重点科技项目(2023ZD008)

申康三年行动计划肺科培育项目(SKPY2021006)