计算机科学 ›› 2024, Vol. 51 ›› Issue (8): 143-151.doi: 10.11896/jsjkx.230700162
徐颖1, 张道强1, 葛荣骏2
XU Ying1, ZHANG Daoqiang1, GE Rongjun2
摘要： 低剂量CT(Low-dose CT,LDCT)扫描在临床实践中起着关键作用,其能有效降低放射科医生和患者的患癌风险。然而,低剂量射线的使用会给生成的CT图像引入明显的噪声,这一问题凸显了LDCT降噪重建的必要性。图像重建领域中的另一个重要任务是超分辨率(Super-resolution,SR)重建,其目标是在减少计算开销的同时实现高分辨率的CT成像。高分辨率CT图像能够更准确地捕捉复杂的解剖细节。尽管这些任务在各自领域取得了显著进展,但目前仍缺乏能够有效利用这两个任务之间固有相关性并同时处理它们的有效方法。文中将边缘信息作为两个任务之间的纽带,并利用梯度提取强相关特征。这使得LDCT降噪重建过程能够辅助超分辨率重建过程,并最终生成具有清晰边缘的结果图像。文中提出的降噪和超分辨率重建网络(NRSR-Net)包括3个组成部分:1)边缘增强框架,该框架利用梯度信息引导和提取相关特征,从而充分利用两个任务之间的相关性,使降噪任务能够辅助超分辨率任务实现更好的性能;2)梯度门控融合模块(Gradient Guided Fusion Block,GGFB),该模块增强高度相关的边缘特征并抑制无关特征,从而实现边缘区域的有效重建;3)梯度损失,该损失函数为模型引入更加丰富的梯度特征,并使网络重点还原边缘区域。一系列的实验表明,NRSR-Net在定量评估中取得了令人满意的峰值信噪比(PSNR)、结构相似性(SSIM)和学习感知图像块相似度(LPIPS),并获得了高质量的可视化结果。这些优势表明NRSR-Net在临床CT成像中具有巨大潜力。
中图分类号:
| [1]SUN F Q,CUI Z Q,ZOU P,et al.Brain Tumor Segmentation Algorithm Based on Multi-scale Features[J].Computer Science,2022,49(S1):12-16. [2]YAO Z W,YANG F,HUANG J,et al.Improved CycleGANs for Intravascular Ultrasound Image Enhancement[J].Computer Science,2019,46(5):221-227. [3]LIU P P,ZHANG W H,LU Z T,et al.Prediction of Malignant and Benign Gastrointestinal Stromal Tumors Based on Radiomics Feature[J].Computer Science,2019,46(1):285-290. [4]WANG L F,WANG R F,LIN S Z,et al.Multimodal Medical Image Fusion Based on Dual Residual Hyper Densely Networks[J].Computer Science,2021,48(2):160-166. [5]GHANI M U,KARL W C.CNN based sinogram denoising for low-dose CT[C]//Mathematics in Imaging.Optica Publishing Group,2018. [6]MA Y J,REN Y,FENG P,et al.Sinogram denoising via attention residual dense convolutional neural network for low-dose computed tomography[J].Nuclear Science and Techniques,2021,32:1-14. [7]CHEN H,ZHANG Y,KALRA M K,et al.Low-dose CT with a residual encoder-decoder convolutional neural network[J].IEEE Transactions on Medical Imaging,2017,36(12):2524-2535. [8]ZHANG Y,HU D,ZHAO Q,et al.CLEAR:comprehensivelearning enabled adversarial reconstruction for subtle structure enhanced low-dose CT imaging[J].IEEE Transactions on Me-dical Imaging,2021,40(11):3089-3101. [9]GE R,HE Y,XIA C,et al.DDPNet:A Novel Dual-Domain Parallel Network for Low-Dose CT Reconstruction[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention.Cham:Springer Nature Switzerland,2022:748-757. [10]YOU C,LI G,ZHANG Y,et al.CT super-resolution GAN constrained by the identical,residual,and cycle learning ensemble(GAN-CIRCLE)[J].IEEE Transactions on Medical Imaging,2019,39(1):188-203. [11]PENG C,LIN W A,LIAO H,et al.Saint:spatially aware interpolation network for medical slice synthesis[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:7750-7759. [12]ZHANG Y,LI K,LI K,et al.MR image super-resolution with squeeze and excitation reasoning attention network[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2021:13425-13434. [13]LI G,LYU J,WANG C,et al.WavTrans:Synergizing wavelet and cross-attention transformer for multi-contrast MRI super-resolution[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention.Cham:Springer Nature Switzerland,2022:463-473. [14]SINGH A,PORIKLI F,AHUJA N.Super-resolving noisy images[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2014:2846-2853. [15]YIM D,KIM B,LEE S.Simultaneous denoising and spatial resolution enhancement using convolutional neural network-based linear model in diagnostic CT images[C]//Medical Imaging 2020:Physics of Medical Imaging.SPIE,2020:1031-1036. [16]HOU H,JIN Q,ZHANG G,et al.CT image quality enhancement via a dual-channel neural network with jointing denoising and super-resolution[J].Neurocomputing,2022,492:343-352. [17]VASWANI A,SHAZEER N,PARMAR N,et al.Attention isall you need[C]//Advances in Neural NIPS’17:Proceedings of the 31st International Conference on Neural Information Processing Systems.2017:6000-6010. [18]WANG Z,CUN X,BAO J,et al.Uformer:A general u-shaped transformer for image restoration[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:17683-17693. [19]WANG X,YU K,WU S,et al.Esrgan:Enhanced super-resolution generative adversarial networks[C]//Proceedings of the European Conference on Computer Vision(ECCV)Workshops.2018. [20]YU L,SHIUNG M,JONDAL D,et al.Development and validation of a practical lower-dose-simulation tool for optimizing computed tomography scan protocols[J].Journal of computer Assisted Tomography,2012,36(4):477-487. [21]YANG L,LI Z,GE R,et al.Low-dose ct denoising via sinogram inner-structure transformer[J].IEEE Transactions on Medical Imaging,2022,42(4):910-921. [22]LEDIG C,THEIS L,HUSZÁR F,et al.Photo-realistic singleimage super-resolution using a generative adversarial network[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:4681-4690. [23]MA C,RAO Y,CHENG Y,et al.Structure-preserving super resolution with gradient guidance[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:7769-7778. [24]LIANG J,CAO J,SUN G,et al.Swinir:Image restoration using swin transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2021:1833-1844. [25]KINGMA D P,BA J.Adam:A method for stochastic optimization[J].arXiv:1412.6980,2014. [26]ZHANG R,ISOLA P,EFROS A A,et al.The unreasonable effectiveness of deep features as a perceptual metric[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:586-595. |
|
||
首页