针对X线图像超分辨率重建的轻量残差注意力网络

doi:10.3969/j.issn.1000-1565.2025.04.010

摘要/Abstract

摘要： 针对当前医学图像超分辨率重建算法复杂、参数量大等问题,提出了轻量化的X线医学图像超分辨率网络LDRAN(lightweight deep residual attention network).该方法设计了轻量且高效的残差块 LDRB(lightweight deep residual block),在保证参数量不增加的条件下,通过增设卷积层来提取更为丰富的图像特征.为进一步提高卷积层间的信息传递效率,设计了一种新颖的残差级联方案IRSC(improved residual skip concatenation).同时,为应对医学影像中信噪比低的问题,构建了多维混合注意力机制模块CSPMA(channel-spatial-pixel mixed attention),该模块分别从通道、空间和像素3个维度筛选信息,从而显著增强了网络对关键图像特征的捕捉能力.实验结果表明,LDRAN在X线医学图像数据集Chest X-ray上的PSNR为36.81 dB,SSIM为0.896 6,均取得了最优.并且能够更好地重建X线图像的细节和纹理.此外,LDRAN在3个自然图像数据集上的重建效果比多数具有代表性的算法更好.

关键词: 超分辨重建, 轻量化, 深度残差块, 混合多维度注意力模块, 残差级联

Abstract: In response to the problem that most current medical image super-resolution reconstruction algorithms are complex and have a large number of parameters, a new lightweight deep residual medical image super-resolution network LDRAN was proposed. By designing a lightweight and efficient residual block LDRB, more convolutional layers are used to extract image information without increasing the number- DOI:10.3969/j.issn.1000-1565.2025.04.010针对X线图像超分辨率重建的轻量残差注意力网络杨昆^1,2,3,齐晁仪⁴,刘天军⁴,艾尚璞¹,闫森广¹,刘秀玲^4,5,薛林雁^1,2,3(1.河北大学质量技术监督学院,河北保定 071002;2.计量仪器与系统国家地方联合工程研究中心,河北保定 071002;3.河北省新能源汽车动力系统轻量化技术创新中心,河北保定 071002;4.河北大学电子信息工程学院,河北保定 071002;5.河北省数字医疗工程重点实验室,河北保定 071002)摘要:针对当前医学图像超分辨率重建算法复杂、参数量大等问题,提出了轻量化的X线医学图像超分辨率网络LDRAN(lightweight deep residual attention network).该方法设计了轻量且高效的残差块 LDRB(lightweight deep residual block),在保证参数量不增加的条件下,通过增设卷积层来提取更为丰富的图像特征.为进一步提高卷积层间的信息传递效率,设计了一种新颖的残差级联方案IRSC(improved residual skip concatenation).同时,为应对医学影像中信噪比低的问题,构建了多维混合注意力机制模块CSPMA(channel-spatial-pixel mixed attention),该模块分别从通道、空间和像素3个维度筛选信息,从而显著增强了网络对关键图像特征的捕捉能力.实验结果表明,LDRAN在X线医学图像数据集Chest X-ray上的PSNR为36.81 dB,SSIM为0.896 6,均取得了最优.并且能够更好地重建X线图像的细节和纹理.此外,LDRAN在3个自然图像数据集上的重建效果比多数具有代表性的算法更好.关键词:超分辨重建;轻量化;深度残差块;混合多维度注意力模块;残差级联中图分类号:TP391 文献标志码:A 文章编号:1000-1565(2025)04-0419-12Super resolution reconstruction of X-ray image based on light residual attention networkYANG Kun^1,2,3, QI Chaoyi⁴, LIU Tianjun⁴, AI Shangpu¹, YAN Senguang¹, LIU Xiuling^4,5, XUE Linyan^1,2,3(1. College of Quality and Technical Supervision, Hebei University, Baoding 071002, China;2. National & Local Joint Engineering Research Center of Metrology Instrument and System, Baoding 071002, China;3. Innovation Center for Lightweight of New Energy Vehicle Power System of Hebei, Baoding 071002, China;4. College of Electronic and Information Engineering, Hebei University, Baoding 071002, China;5. Key Laboratory of Digital Medical Engineering of Hebei Province, Baoding 071002, China)Abstract: In response to the problem that most current medical image super-resolution reconstruction algorithms are complex and have a large number of parameters, a new lightweight deep residual medical image super-resolution network LDRAN was proposed. By designing a lightweight and efficient residual block LDRB, more convolutional layers are used to extract image information without increasing the number- 收稿日期:2024-11-21;修回日期:2025-03-18 基金项目:河北省自然科学基金项目(F2022201037);河北大学科研创新团队项目(IT2023B07) 第一作者:杨昆(1976—),男,河北大学教授,博士生导师,主要从事生物医学图像处理方向研究.E-mail: yangkun@hbu.edu.cn 通信作者:薛林雁(1981—),女,河北大学教授,主要从事生物医学图像处理方向研究.E-mail:lyxue@hbu.edu.cn 第4期杨昆等:针对X线图像超分辨率重建的轻量残差注意力网络河北大学学报(自然科学版) 第45卷of parameters, and a new residual cascade method IRSC(improved residual skip concatenation)is designed to improve the information utilization among convolutional layers. In addition, to solve the problem of small percentage of useful information in medical images, the improved mixed multidimensional attention module CSPMA(channel-spatial-pixel mixed attention)is designed to filter information from three dimensions: channel, spatial and pixel, and is used to improve the networks attention to useful information in images. The experimental results show that LDRAN achieves the best PSNR(36.81 dB)and SSIM(0.8966)on the X-ray medical image dataset Chest X-ray and can better reconstruct the details and textures of X-ray images. Additionally, LDRAN outperforms most representative reconstruction algorithm on three natural image datasets.

Key words: super-resolution reconstruction, lightweight, deep residual block, mixed multidimensional attention module, residual cascade

中图分类号:

TP391

杨昆,齐晁仪,刘天军,艾尚璞,闫森广,刘秀玲,薛林雁. 针对X线图像超分辨率重建的轻量残差注意力网络[J]. 河北大学学报(自然科学版), 2025, 45(4): 419-430.

YANG Kun, QI Chaoyi, LIU Tianjun, AI Shangpu, YAN Senguang, LIU Xiuling, XUE Linyan. Super resolution reconstruction of X-ray image based on light residual attention network[J]. Journal of Hebei University(Natural Science Edition), 2025, 45(4): 419-430.

参考文献

[1] 杨昆,罗萍,吕一品,等.一种基于 GA_Faster R-CNN 的掌指骨骨折计算机辅助诊断系统[J].河北大学学报(自然科学版), 2021, 41(4): 412. DOI: 10.3969/j.issn.1000-1565.2021.04.011.
[2] 刘爽,田兆星,李浩然,等.一种基于改进 YOLOv5s 网络的结直肠腺瘤实时检测方法[J].河北大学学报(自然科学版), 2022, 42(3): 327. DOI: 10.3969/j.issn.1000-1565.2022.03.016.
[3] 李浩然,刘琨,常世龙,等.基于残差混合域注意力网络的PET超分辨率重建方法[J].电子测量技术, 2021, 44(14): 103-110. DOI:10.19651/j.cnki.emt.2106998.
[4] 薛林雁,李轩昂,齐晁仪,等.基于改进 YOLOv5s 的肠镜息肉多分类实时检测方法[J].河北大学学报(自然科学版), 2024, 44(4): 424. DOI: 10.3969/j.issn.1000-1565.2024.04.011
[5] PINEDA F, AYMA V, BELTRAN C. A generative adversarial network approach for super-resolution of sentinel-2 satellite images[J]. Int Arch Photogramm Remote Sens Spatial Inf Sci, 2020, XLIII-B1-2020: 9-14. DOI:10.5194/isprs-archives-xliii-b1-2020-9-2020.
[6] 许博鸣,刘晓峰,业巧林,等.面向移动平台的深度学习复杂场景目标识别应用[J].陕西师范大学学报(自然科学版), 2019, 47(5): 10-15. DOI:10.15983/j.cnki.jsnu.2019.05.152.
[7] 陈晓,荆茹韵.单图像超分辨率方法综述[J].电子测量技术, 2022, 45(9): 104-112. DOI:10.19651/j.cnki.emt.2208826.
[8] KEYS R. Cubic convolution interpolation for digital image processing[J]. IEEE Trans Acoust Speech Signal Process, 1981, 29(6): 1153-1160. DOI:10.1109/TASSP.1981.1163711.
[9] DONG C, LOY C C, HE K M, et al. Image super-resolution using deep convolutional networks[J]. IEEE Trans Pattern Anal Mach Intell, 2016, 38(2): 295-307.
[10] DONG C, LOY C C, TANG X O. Accelerating the super-resolution convolutional neural network[M] //Computer Vision-ECCV 2016, Cham: Springer International Publishing, 2016: 391-407. DOI:10.1007/978-3-319-46475-6_25.
[11] KIM J, LEE J K, LEE K M. Accurate image super-resolution using very deep convolutional networks[C] //2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Las Vegas, USA, IEEE, 2016: 1646-1654. DOI:10.1109/CVPR.2016.182.
[12] KIM J, LEE J K, LEE K M. Deeply-recursive convolutional network for image super-resolution[C] //2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Las Vegas, USA, IEEE, 2016: 1637-1645. DOI:10.1109/CVPR.2016.181.
[13] TAI Y, YANG J, LIU X M. Image super-resolution via deep recursive residual network[C] //2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Honolulu, USA, IEEE, 2017: 2790-2798. DOI:10.1109/CVPR.2017.298.
[14] LAI W S, HUANG J B, AHUJA N, et al. Fast and accurate image super-resolution with deep Laplacian pyramid networks[J]. IEEE Trans Pattern Anal Mach Intell, 2019, 41(11): 2599-2613. DOI:10.1109/TPAMI.2018.2865304.
[15] CHU X J, CHEN L Y, YU W Q. NAFSSR: stereo image super-resolution using NAFNet[C] //2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops(CVPRW), New Orleans, USA, IEEE, 2022: 1238-1247. DOI:10.1109/CVPRW56347.2022.00130.
[16] QIU D F, ZHENG L X, ZHU J Q, et al. Multiple improved residual networks for medical image super-resolution[J]. Future Generation Computer Systems, 2021, 116: 200-208. DOI:10.1016/j.future.2020.11.001.
[17] DU Y B, JIA R S, CUI Z, et al. X-ray image super-resolution reconstruction based on a multiple distillation feedback network[J]. Applied Intelligence, 2021, 51(7): 5081-5094. DOI:10.1007/s10489-020-02123-2.
[18] CHEN Y H, ZHENG Q Y, CHEN J S. Double paths network with residual information distillation for improving lung CT image super resolution[J]. Biomed Signal Process Contr, 2022, 73: 103412. DOI:10.1016/j.bspc.2021.103412.
[19] SHI W Z, CABALLERO J, HUSZÁR F, et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network[C] //2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Las Vegas, USA, IEEE, 2016: 1874-1883. DOI:10.1109/CVPR.2016.207.
[20] 杨昆,原嘉成,高聪,等.基于改进的 Faster R-CNN 的息肉目标检测和分类方法[J].河北大学学报(自然科学版), 2023, 43(1): 103. DOI: 10.3969/j.issn.1000-1565.2023.01.015.
[21] AHN N, KANG B, SOHN K A. Fast, accurate, and lightweight super-resolution with cascading residual network[C] // Computer Vision-ECCV 2018, Cham: Springer International Publishing, 2018: 256-272. DOI:10.1007/978-3-030-01249-6_16.
[22] ZHANG Y L, TIAN Y P, KONG Y, et al. Residual dense network for image super-resolution[C] //2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, IEEE, 2018: 2472-2481. DOI:10.1109/CVPR.2018.00262.
[23] PENG C M, SHU P, HUANG X Y, et al. LCRCA: image super-resolution using lightweight concatenated residual channel attention networks[J]. ApplIntell, 2022, 52(9): 10045-10059. DOI:10.1007/s10489-021-02891-5.
[24] LI X, WANG W H, HU X L, et al. Selective kernel networks[C] //2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), Long Beach, USA, IEEE, 2019: 510-519. DOI:10.1109/cvpr.2019.00060.
[25] WANG X S, PENG Y F, LU L, et al. ChestX-Ray8: hospital-scale chest X-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases[C] //2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Honolulu, USA, IEEE, 2017: 3462-3471. DOI:10.1109/CVPR.2017.369.
[26] YUAN F, HUANG L F, YAO Y. An improved PSNR algorithm for objective video quality evaluation[C] //2007 Chinese Control Conference, Zhangjiajie, China, IEEE, 2007: 376-380. DOI:10.1109/CHICC.2006.4347144.
[27] WANG Z, BOVIK A C, SHEIKH H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Trans Image Process, 2004, 13(4): 600-612. DOI:10.1109/TIP.2003.819861.
[28] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C] //Computer Vision – ECCV 2018, ACM, 2018: 3-19. DOI:10.1007/978-3-030-01234-2_1.
[29] ZHAO H Y, KONG X T, HE J W, et al. Efficient image super-resolution using pixel attention[M] //Computer Vision-ECCV 2020 Workshops, Cham: Springer International Publishing, 2020: 56-72. DOI:10.1007/978-3-030-67070-2_3.
[30] LIM B, SON S, KIM H, et al. Enhanced deep residual networks for single image super-resolution[C] //Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops,2017.abs/1707.02921:1132-1140.DOI:10.1109/CUPRW.2017.151.
[31] HUI Z, GAO X, YANG Y, et al. Lightweight image super-resolution with information multi-distillation net-work[C] //Proceedings of the 27th ACM International Conference on Multimedia,2019,abs/1909.11856:2024-2032.DOI: 10.1145/3343031.3351084. (