基于全卷积神经网络的低剂量CT去噪算法

针对低剂量医学CT图像因减少辐射剂量而引入大量噪声，导致图像质量显著下降的问题，提出一种基于残差注意力机制和自适应特征融合的低剂量CT图像去噪算法，该算法使用全卷积神经网络来完成低剂量CT图像去噪。在网络框架中引入残差注意力机制和选择性内核特征融合模块，以过滤噪声信息，提取有效特征并自适应地融合图像特征，避免重建过程中的细节损失，提高图像质量，使去噪后的图像在感知上更接近原始图像。定性和定量实验表明，与现有的算法相比，在真实的临床数据集上，所提出的算法能够有效地抑制噪声，并恢复低剂量CT图像中更详细的纹理。与低剂量CT图像相比，所提出的算法将峰值信噪比提高14.94%，结构相似度提高4.68%，均方根误差降低40.11%，可以满足医学成像的诊断要求。     

一种医学图像去噪程序的并行优化

目的：在采集、处理和传输过程中,医学图像会存在各种噪声,严重影响医学图像的质量和后续对图像的各种处理,因此医学图像去噪具有重要意义。同时医学图像数据量大,去噪处理算法复杂,在一般个人电脑上进行医学图像去噪仍是一个非常耗时的过程．很难满足实际应用中高实时性的要求．因此需要通过优化来提高去噪的处理速度。方法：本文利用CUDA（Compute Unified Device Architecture）并行编程对基于同质算法的三维医学图像去噪进行加速,CPU和GPU（Graphic ProcessorUnit）异构编程方式能发挥GPU高强度的计算能力,提高算法的执行速度。通过使用纹理存储器将图像数据与纹理绑定,优化存储器访问,提高数据访问速度。优化过程中,合理选择三维图像数据的分块方式和线程块维度。可以获得更快的加速。结果：与基于同质的matlab和CPU去噪程序相比,并行优化后GPU程序在保持去噪图像质量的前提卞可以达到几百倍的加速。结论：CUDA加速大大缩短了三维医学图像去噪的运行时间,解决了医学图像去噪的速度瓶颈问题．可以应用于对运行速度有要求的图像处理中。     

基于卷积神经网络检测颈动脉斑块

提出一种基于卷积神经网络的自动检测超声图像颈动脉斑块的方法。通过超分辨生成对抗网络提高超声图像质量,并采用高斯混合模型算法结合先验知识自动提取感兴趣区域;最后采用卷积神经网络实现颈动脉有无斑块的自动检测。使用上海市奉贤区中心医院提供的数据集,自动检测颈动脉是否有斑块,模型准确度、敏感度、特异度分别达到94.11%、96.30%、91.67%。实验证明基于卷积神经网络检测颈动脉斑块结果和真实值有很高的一致性,且鲁棒性好。     

基于变分法的超声图像斑点噪声自适应滤波算法

各向异性扩散模型在去除超声图像斑点噪声时不能有效保护图像细节,针对上述问题本文提出基于变分法的自适应最小能量去噪模型.首先直接将由微分方程表示的各向异性扩散模型转化为最小能量变分模型;然后引入欧拉弹性能量模型,在去除噪声的同时有效地保护和增强图像细节.同时为了解决数值求解过程中出现的迭代次数与迭代步长的矛盾,本文还提出迭代停止准则和自适应变步长去噪算法.仿真和真实超声图像的实验结果表明基于变分法的超声图像斑点噪声自适应滤波算法在去噪的同时能够很好地保护细节信息,而且能有效地减少迭代次数.     

基于小波统计模型的医学超声图像去噪方法研究

超声图像中固有的斑点噪声严重降低图像的可解译程度,影响了后续的图像分析和诊断.因此,抑制相干斑噪声一直是医学超声图像预处理中一个关键性问题.本研究通过对含斑图像做对数变换和冗余小波分解,提出了一种基于Bayesian估计的小波域局部自适应性去斑算法.将斑点噪声和有用信号的小波系数分别建模为瑞利分布和拉普拉斯分布,利用最大后验概率(MAP)准则得到了一种解析的Bayesian估计表达式;进一步通过邻域窗口估计模型参数,使算法具有局部自适应性.实验仿真表明,该算法简单有效,在滤除超声图像斑点噪声的同时,较好地保持了图像的细节特征,其性能优于空间域滤波和传统的小波去噪算法.     

基于深度学习的低剂量CT图像去噪

目的:提出一种基于深度学习的方法用于低剂量CT（LDCT）图像的噪声去除。方法:首先进行滤波反投影重建,然后利用多尺度并行残差U-net（MPR U-net）的深度学习模型对重建后的LDCT图像进行去噪。实验数据采用LoDoPaB-CT挑战赛的医学CT数据集,其中训练集35 820张图像,验证集3 522张图像,测试集3 553张图像,并采用峰值信噪比（PSNR）与结构相似性系数（SSIM）来评估模型的去噪效果。结果:LDCT图像处理前后PSNR分别为28.80、38.22 dB,SSIM分别为0.786、0.966,平均处理时间为0.03 s。结论:MPR U-net深度学习模型能较好地去除LDCT图像噪声,提升PSNR,保留更多图像细节。     

非局部主成分分析极大似然估计MRI图像Rician噪声去噪

由于MRI图像中噪声呈Rician分布,直接使用现有针对高斯噪声的去噪方法将引入误差。基于此本研究使用Rician噪声模型改进现有极大似然估计去噪的高斯模型,同时引入非局部主成分分析,在非局部区域选择灰度和纹理均具有较高相似性的像素进行最优复原估计。使用非局部主成分分析不仅克服现有局部性去噪方法模糊边界的缺陷,而且具有更高的图像细节信息复原能力。分别应用所提出的方法、局部极大似然估计去除Rician噪声方法、采用参数修正非局部均值去除Rician噪声方法、无特定噪声模型的全变差方法,对不同噪声等级和不同纹理复杂度的图像进行定性和定量的去噪实验。结果表明,所提出的方法可在保持图像细节和纹理信息的前提下有效去噪,较之现有方法效果更好。     

基于形态学重构的超声医学图像滤波方法

Speckle噪声是造成超声医学图像质量下降的最主要原因。我们通过修改形态学重构算法-Downhill算法的初始条件,使其适用于超声医学图像的去噪处理。首先在掩模图像中确定标记图像作为算法的初始化和开始区域,再使用改进的Downhill算法对超声医学图像进行滤波处理。实验结果表明,与其他3种传统滤波方法相比,该方法能快速有效地去除心脏腔室内的Speckle噪声同时保留图像的轮廓细节信息。     

基于Prewitt算法的颅内CT图像病灶分割算法

针对颅内CT图像病灶周围存在大量噪声,分割结果欠佳的问题,本研究提出基于Prewitt算法的颅内CT图像病灶分割算法。首先采用改进型中值小波去噪算法,去除颅内CT图像中的噪声点,优化图像质量;然后使用基于Prewitt算法的图像分割法,完成去噪后CT图像的病灶分割。结果表明,本研究算法在分割颅内CT图像病灶时,错分率为对比算法的1/10,并可将颅内CT图像的噪声点全部去除。说明该算法对颅内CT图像病灶的分割可行性,可用于颅内CT图像病灶分割。     

基于稀疏分解的医学CT图像去噪

医学CT图像成像过程中,由于成像机制的影响,不可避免的引入噪声。图像中的噪声会降低图像质量,影响临床诊断。因此,有必要对医学CT图像进行去噪处理。本文采用图像的稀疏分解方法来对混有噪声的肝癌CT图像进行消噪处理,提出分块稀疏分解去噪。实验表明,本文算法对医学图像中噪声去除有一定效果。在分解原子个数相同的条件下,本文方法去噪后重建图像比在整幅图像上进行稀疏去噪重建的计算速度提高了约15倍。     

基于比率距离的自适应超声图像去噪方法

针对超声图像噪声的瑞利分布特性,使用一种新的自适应超声图像去噪方法,改进固定窗口包含边缘时无法做到沿边缘方向滤波的不足。采用可自由伸缩的自适应滤波窗口,首先针对瑞利分布的噪声引入比率距离,得到超声图像像素间的相似度距离,然后考虑像素的邻域图像块均值,解决相似度距离之间比较的问题,最后像素根据新的相似度距离进行八方向伸展,得到不规则形状的滤波窗口进行去噪。用仿真超声图像和临床超声图像进行实验,图像评价指标结果表明该算法优于经典算法,更适用于去除超声图像的斑点噪声,在去除噪声的同时能够较好地保留细节边缘。     

Robust non-homomorphic approach for speckle reduction in medical ultrasound images

Most existing wavelet-based image denoising techniques are developed for additive white Gaussian noise. In applications to speckle reduction in medical ultrasound (US) images, the traditional approach is first to perform the logarithmic transform (homomorphic processing) to convert the multiplicative speckle noise model to an additive one, and then the wavelet filtering is performed on the log-transformed image, followed by an exponential operation. However, this non-linear operation leads to biased estimation of the signal and increases the computational complexity of the filtering method. To overcome these drawbacks, an efficient, non-homomorphic technique for speckle reduction in medical US images is proposed. The method relies on the true characterisation of the marginal statistics of the signal and speckle wavelet coefficients. The speckle component was modelled using the generalised Nakagami distribution, which is versatile enough to model the speckle statistics under various scattering conditions of interest in medical US images. By combining this speckle model with the generalised Gaussian signal first, the Bayesian shrinkage functions were derived using the maximum a posteriori (MAP) criterion. The resulting Bayesian processor used the local image statistics to achieve soft-adaptation from homogeneous to highly heterogeneous areas. Finally, the results showed that the proposed method, named GNDShrink, yielded a signal-to-noise ratio (SNR) gain of 0.42 dB over the best state-of-the-art despeckling method reported in the literature, 1.73 dB over the Lee filter and 1.31 dB over the Kaun filter at an input SNR of 12.0 dB, when tested on a US image. Further, the visual comparison of despeckled US images indicated that the new method suppressed the speckle noise well, while preserving the texture and organ surfaces.     

基于多尺度非线性扩散的超声图像斑点噪声去除算法

针对现有去噪算法可能造成超声图像细节模糊甚至丢失的问题,本文提出基于多尺度非线性扩散（multiscale nonlinear diffusion,MSND）的超声图像去噪模型.该模型结合冗余拉普拉斯塔形数据分解和非线性扩散的优点,利用冗余拉普拉斯塔形数据分解将图像分解为等大小的空间-频率子带,综合各子带的特征得到图像边缘和细节的精细表示,然后根据所得的综合特征指导各子带图像的非线性扩散.实验结果表明本文算法在去除噪声的同时能有效地保留和增强边界与细节.     

Homomorphic wavelet thresholding technique for denoising medical ultrasound images

A novel homomorphic wavelet thresholding technique for reducing speckle noise in medical ultrasound images is presented. First, we show that the speckle wavelet coefficients in the logarithmically transformed ultrasound images are best described by the Nakagami family of distributions. By exploiting this speckle model and the Laplacian signal prior, a closed form, data-driven, and spatially adaptive threshold is derived in the Bayesian framework. The spatial adaptivity allows the additional information of the image (such as identification of homogeneous or heterogeneous regions) to be incorporated into the algorithm. Further, the threshold has been extended to the redundant wavelet representation, which yields better results than the decimated wavelet transform. Experimental results demonstrate the improved performance of the proposed method over other well-known speckle reduction filters. The application of the proposed method to a realistic US test image shows that the new technique, named HomoGenThresh, outperforms the best wavelet-based denoising method reported in [1] by more than 1.6 dB, Lee filter by 3.6 dB, Kaun filter by 3.1 dB and band-adaptive soft thresholding [2] by 2.1 dB at an input signal-to-noise ratio (SNR) of 13.6 dB.     

Digital Radiographic Image Denoising Via Wavelet-Based Hidden Markov Model Estimation

This paper presents a technique for denoising digital radiographic images based upon the wavelet-domain Hidden Markov tree (HMT) model. The method uses the Anscombes transformation to adjust the original image, corrupted by Poisson noise, to a Gaussian noise model. The image is then decomposed in different subbands of frequency and orientation responses using the dual-tree complex wavelet transform, and the HMT is used to model the marginal distribution of the wavelet coefficients. Two different correction functions were used to shrink the wavelet coefficients. Finally, the modified wavelet coefficients are transformed back into the original domain to get the denoised image. Fifteen radiographic images of extremities along with images of a hand, a line-pair, and contrast–detail phantoms were analyzed. Quantitative and qualitative assessment showed that the proposed algorithm outperforms the traditional Gaussian filter in terms of noise reduction, quality of details, and bone sharpness. In some images, the proposed algorithm introduced some undesirable artifacts near the edges.     

Off-line determination of the optimal number of iterations of the robust anisotropic diffusion filter applied to denoising of brain MR images

Although anisotropic diffusion filters have been used extensively and with great success in medical image denoising, one limitation of this iterative approach, when used on fully automatic medical image processing schemes, is that the quality of the resulting denoised image is highly dependent on the number of iterations of the algorithm. Using many iterations may excessively blur the edges of the anatomical structures, while a few may not be enough to remove the undesirable noise. In this work, a mathematical model is proposed to automatically determine the number of iterations of the robust anisotropic diffusion filter applied to the problem of denoising three common human brain magnetic resonance (MR) images (T1-weighted, T2-weighted and proton density). The model is determined off-line by means of the maximization of the mean structural similarity index, which is used in this work as metric for quantitative assessment of the resulting processed images obtained after each iteration of the algorithm. After determining the model parameters, the optimal number of iterations of the algorithm is easily determined without requiring any extra computation time. The proposed method was tested on 3D synthetic and clinical human brain MR images and the results of qualitative and quantitative evaluation have shown its effectiveness.     

An Effective Non-rigid Registration Approach for Ultrasound Image Based On “Demons” Algorithm

Medical image registration is an important component of computer-aided diagnosis system in diagnostics, therapy planning, and guidance of surgery. Because of its low signal/noise ratio (SNR), ultrasound (US) image registration is a difficult task. In this paper, a fully automatic non-rigid image registration algorithm based on demons algorithm is proposed for registration of ultrasound images. In the proposed method, an “inertia force” derived from the local motion trend of pixels in a Moore neighborhood system is produced and integrated into optical flow equation to estimate the demons force, which is helpful to handle the speckle noise and preserve the geometric continuity of US images. In the experiment, a series of US images and several similarity measure metrics are utilized for evaluating the performance. The experimental results demonstrate that the proposed method can register ultrasound images efficiently, robust to noise, quickly and automatically.     

A Motion Compounding Technique for Speckle Reduction in Ultrasound Images

The quality of ultrasound images is usually influenced by speckle noise and the temporal decorrelation of the speckle patterns. To reduce the speckle noise, compounding techniques have been widely applied. Partially correlated images scanned on the same subject cross-section are combined to generate a compound image with improved image quality. However, the compounding technique might introduce image blurring if the transducer or the target moves too fast. This blurring effect becomes especially critical when assessing tissue deformation in clinical motion examinations. In this paper, an ultrasound motion compounding system is proposed to improve the quality of ultrasound motion sequences. The proposed motion compounding technique uses a hierarchical adaptive feature weighted motion estimation method to realign the frames before compounding. Each frame is first registered and warped to the reference frame before being compounded to reduce the speckle noise. Experimental results showed that the motion could be assessed accurately and better visualization could be achieved for the compound images, with improved signal-to-noise and contrast-to-noise ratios.     

Nonlocal total variation based on symmetric Kullback-Leibler divergence for the ultrasound image despeckling

Background

Ultrasound imaging is safer than other imaging modalities, because it is noninvasive and nonradiative. Speckle noise degrades the quality of ultrasound images and has negative effects on visual perception and diagnostic operations.

Methods

In this paper, a nonlocal total variation (NLTV) method for ultrasonic speckle reduction is proposed. A spatiogram similarity measurement is introduced for the similarity calculation between image patches. It is based on symmetric Kullback-Leibler (KL) divergence and signal-dependent speckle model for log-compressed ultrasound images. Each patch is regarded as a spatiogram, and the spatial distribution of each bin of the spatiogram is regarded as a weighted Gamma distribution. The similarity between the corresponding bins of the two spatiograms is computed by the symmetric KL divergence. The Split-Bregman fast algorithm is then used to solve the adapted NLTV object function. Kolmogorov-Smirnov (KS) test is performed on synthetic noisy images and real ultrasound images.

Results

We validate our method on synthetic noisy images and clinical ultrasound images. Three measures are adopted for the quantitative evaluation of the despeckling performance: the signal-to-noise ratio (SNR), structural similarity index (SSIM), and natural image quality evaluator (NIQE). For synthetic noisy images, when the noise level increases, the proposed algorithm achieves slightly higher SNRS than that of the other two algorithms, and the SSIMS yielded by the proposed algorithm is obviously higher than that of the other two algorithms. For liver, IVUS and 3DUS images, the NIQE values are 8.25, 6.42 and 9.01, all of which are higher than that of the other two algorithms.

Conclusions

The results of the experiments over synthetic and real ultrasound images demonstrate that the proposed method outperforms current state-of-the-art despeckling methods with respect to speckle reduction and tissue texture preservation.