基于双流Transformer结构的多能计算机断层扫描成像投影数据去噪方法  

Projection Domain Denoising Method for Multi-Energy Computed Tomography via Dual-Stream Transformer

作　　者：欧阳顺馨 史再峰 孔凡宁 张丽丽 曹清洁[2] Ouyang Shunxin;Shi Zaifeng;Kong Fanning;Zhang Lili;Cao Qingjie(School of Microelectronics,Tianjin University,Tianjin 300072,China;College of Mathematical Science,Tianjin Normal University,Tianjin 300387,China;Tianjin Key Laboratory of Imaging and Sensing Microelectronics Technology,Tianjin 300072,China)

机构地区：[1]天津大学微电子学院,天津300072 [2]天津师范大学数学科学学院,天津300387 [3]天津市成像与感知微电子技术重点实验室,天津300072

出　　处：《激光与光电子学进展》2024年第8期399-408,共10页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(62071326)。

摘　　要：多能计算机断层扫描(Computed tomography,CT)技术可以更加精确地分辨出人体组织对不同能量X射线光子的吸收情况,是医学成像领域的重要发展方向。为了解决因量子噪声等非理想效应加重导致重建图像质量急剧退化的问题,提出了一种基于移位窗口多头自注意力机制的双流Transformer网络结构。该结构利用移位窗口Transformer和局部增强窗口Transformer分别提取投影数据的全局和局部特征,充分利用投影数据的非局部自相似性以保留投影数据的内部结构;然后通过残差卷积融合提取的特征;最后使用带有非局部全变分的混合损失函数来监督网络模型的训练,提升该网络模型对投影数据内部细节的敏感程度。实验结果表明,所提方法处理后的重建图像峰值信噪比(PSNR)值、结构相似性(SSIM)值和特征相似度(FSIM)值分别达到37.7301 dB、0.9944和0.9961。与目前先进的多能CT去噪方法相比,所提方法在去除低剂量多能CT投影数据噪声的同时,可保留更多的细节特征,有利于后续的精确诊断。The multienergy computed tomography(CT)technique can resolve the absorption rates of various energy Xray photons in human tissues,representing a significant advancement in medical imaging.By addressing the challenge of swift degradation in reconstructed image quality,primarily due to nonideal effects such as quantum noise,a dualstream Transformer network structure is introduced.This structure utilises the shiftedwindow multihead selfattention denoising approach for projection data.The shifted windows Transformer extracts the global features of the projection data,while the locallyenhanced window Transformer focuses on local features.This dual approach capitalizes on the nonlocal selfsimilarity of the projection data to maintain its inherent structure,subsequently merged by residual convolution.For model training oversight,a hybrid loss function incorporating nonlocal total variation is employed,which enhances the network model’s sensitivity to the inner details of the projected data.Experimental results demonstrate that our method’s processed projection data achieve a peak signal to noise ratio(PSNR)of 37.7301 dB,structure similarity index measurement(SSIM)of 0.9944,and feature similarity index measurement(FSIM)of 0.9961.Relative to leading denoising techniques,the proposed method excels in noise reduction while preserving more inner features,crucial for subsequent accurate diagnostics.

关 键 词：图像处理 计算机断层扫描成像 低剂量 卷积神经网络 Transformer网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]