基于在线字典学习的医学图像特征提取与融合

提出一种基于在线字典学习(ODL)的医学图像特征提取与融合的新算法。首先,采用大小为8像素×8像素的滑动窗处理源图像,得到联合矩阵;通过ODL算法得到该联合矩阵的冗余字典,并利用最小角回归算法(LARS)计算该联合矩阵的稀疏编码;将稀疏编码列向量的1范数作为稀疏编码的活动级测量准则,然后根据活动级最大准则融合稀疏编码;最后根据融合后的稀疏编码和冗余字典重构融合图像。实验图像为20位患者的已配准脑部CT和MR图像,采用5种性能指标评价融合图像的质量,同两种流行的融合算法比较。结果显示,所提出算法的各项客观指标均值最优,Piella指数、QAB/F指数、MIAB/F指数、BSSIM指数和空间频率的均值分别为0.800 4、0.552 4、3.630 2、0.726 9和31.941 3,融合图像对比度、清晰度高,病灶的边缘清晰,运行速度较快,可以辅助医生诊断和临床治疗。     

基于在线字典学习和脉冲耦合神经网络的脑图像融合

医学图像融合是医学影像和放射医学等领域的研究热点之一,广受医学界和工程界重视。提出一种基于在线字典学习(ODL)和脉冲耦合神经网络(PCNN)的脑部CT和MR图像融合新算法。首先,利用滑动窗技术将源图像分块,使用ODL算法和最小角回归算法(LARS)得到各图像块对应列向量的稀疏编码;其次,将稀疏编码作为脉冲耦合神经网络的外部输入刺激信号进行迭代处理,根据点火次数确定融合系数;最后,根据融合系数和学习字典重构融合图像。基于哈佛医学院的10组脑部CT和MR数据,将所提出算法同基于KSVD的融合算法、基于ODL的融合算法、基于NSCT的融合算法比较。实验结果显示：综合考虑主观视觉效果和客观评价指标,该算法性能整体优于其他算法,客观参数指标BSSIM、MI、Piella、SF、STD、Q^AB/F的均值分别为0.751 2、3.769 6、0.697 1、29.526 7、90.090 6、0.570 7,可以提供丰富的信息来辅助医生分析病变体,提高临床医疗诊断的准确性和治疗规划的科学性。     

联合稀疏表示的医学图像融合及同步去噪

将多模态医学图像的互补信息有机地融合在一起,可为临床诊断和辅助治疗提供丰富信息和有效帮助。基于联合稀疏模型,提出一种联合稀疏表示的医学图像融合算法,当图像被噪声污染时,该算法在融合的同时兼有去噪功能。首先,将配准的源图像编纂成列向量并组成联合矩阵,通过在线字典学习算法(ODL)得到该矩阵的超完备字典;其次,利用该字典得到联合稀疏模型下的联合字典,之后利用最小角回归算法(LARS)计算基于联合字典的公共稀疏系数和各图像的独特稀疏系数,并根据“选择最大化”融合规则得到融合图像的稀疏系数;最后,根据融合系数和超完备字典重构融合图像。将该算法与3种经典算法比较,结果显示其主观上亮度失真和对比度失真较小,边缘纹理清晰,客观参数指标MI、Q^AB/F在无噪声干扰和有噪声干扰时的统计均值分别为：3.992 3、2.896 4、2.505 5和0.658、0.552 4、0.439 6,可以为临床诊断和辅助治疗提供有效帮助。     

基于Contourlet变换和PCNN的CT图像椎体解剖轮廓特征提取方法的研究

提出一种新的基于Contourlet变换和脉冲耦合神经网络(PCNN)的医学图像解剖轮廓特征提取算法。首先对原始椎体CT图像进行Contourlet变换,得到能稀疏表示图像边缘以及方向信息的子带和低频子带;然后结合PCNN对低频子带进行边缘轮廓细节提取,最后利用处理后的所有子带系数,通过Contourlet逆变换,提取出图像的边缘轮廓。实验将本算法提取的结果与Canny算子、区域生长法以及结合小波变换和PCNN的算法提取的图像边缘轮廓进行比较,结果表明新算法能够有效的实现医学图像解剖结构轮廓特征的提取。     

基于双稀疏字典的医学图像融合算法及在脑血管疾病诊断中的应用

目的通过图像融合方法结合解剖和功能医学图像以提供更多有用的信息并辅助医生诊断。方法利用稀疏表示能很好地反映图像特征的优势。首先,选取医院脑梗死和脑出血的CT和MRI的临床图像,采用双稀疏字典算法得到稀疏字典,再通过结合空间域信息的最大选择法作为融合规则对其进行融合,并与基于主成分分析(principal component analysis,PCA)和离散小波变换(discrete wavelet transform,DWT)方法的图像融合结果在主观方面以及客观方面的QAB/F和Piella指标上进行比较。结果本文提出的方法所获得的融合图像主观评价优于另外两种方法。QAB/F和Piella的均值分别为0.9139和0.7213,客观评价指标也优于另外两种方法。结论基于双稀疏字典的图像融合算法得到的融合图像更清晰,对比度更高,并且特征保留效果更好,有助于医生的诊断。     

基于NSST变换和PCNN的医学图像融合方法

目的:融合PET/CT/MRI医学图像,使结果图像尽可能包含更多边缘和纹理特征等信息,以更好地区分病变、肿瘤与正常组织器官,为疾病诊断提供更多的有用信息。方法:提出一种基于非下采样剪切波变换(NSST)和脉冲耦合神经网络(PCNN)模型的融合方法。首先,根据图像局部区域能量和,对图像NSST低频系数进行加权融合;然后,根据PCNN神经元的点火次数,选择图像NSST高频方向系数;最后,通过逆NSST变换,得到融合后的图像。结果:分别对7组MRI/PET和CT/PET图像进行融合实验,其结果图像具有很好的视觉效果,且在互信息、边缘相似性、梯度相似性及空间频率4个指标综合评价中较其它算法更优。结论:本方法可以自适应捕获边缘和纹理信息,具有良好的融合效果。     

基于加权压缩感知的MR图像重建方法

背景：压缩感知理论已广泛应用于MR图像的快速重建中。在对K空间数据进行随机欠采样后,通过非线性优化算法求解带约束的范数最小化问题,可恢复出在变换域具有稀疏性的MR图像。 目的：为了增强图像在变换域中的稀疏性,改善MR图像重建质量,提出了对待重建图像的稀疏表示进行加权的方法。 方法：采用非线性共轭梯度下降算法求解该加权范数最小化问题,在迭代过程中,根据所求取的图像稀疏表示来更新权值矩阵,增强MR图像的稀疏性。 结果与结论：通过比较带加权矩阵和不带加权矩阵的压缩感知图像重建方法,结果表明带加权矩阵改进的算法提高了图像重建能力。     

基于剪切波变换的医学图像融合算法

剪切波变换是一种新颖的多尺度几何分析工具,具有多分辨率、多方向性、效率较高等优点,比小波变换、曲波变换、轮廓波变换等图像表示方法有独特有的优势.基于剪切波变换提出一种医学图像融合算法,先将原始图像通过剪切波变换分解为低频子带图像和高频方向子带图像,然后采用非负矩阵分解方法融合低频子带系数,再通过深入研究人类视觉系统的特性提出最大视觉能量对比度方法,利用局部对比度和局部区域的能量和进行高频方向子带系数的融合,最后通过剪切波逆变换得到融合图像.两组实验均显示所提出的融合方法在与其余3种融合方法的比较中,采用的5项客观评价指标均有4项指标达到最优值,证明所提出的方法获取的融合图像效果最好.     

基于双层字典学习的低剂量CT图像重建算法

目的低剂量投影条件下的CT图像重建。方法采用双层K-奇异值分解(K-singular value decomposition,K-SVD)字典训练的学习方法进行图像的超分辨率重建。字典学习方法中采用KSVD算法,稀疏编码采用正交匹配追踪(orthogonal matching pursuit,OMP)算法。该算法首先利用训练库进行第一层字典训练,然后利用第一层训练的字典对低分辨率图像进行重建。进而将重建图像作为第二层待重建图像的输入,这样使得第二层输入图像含有较多的高频细节信息,因此能在重构的过程中恢复更多的细节信息,让高分辨率重构图像达到较好的效果。结果双层字典重建效果明显优于KSVD算法,重建图像更接近于原始高分辨率CT图像。结论本研究对双层字典训练学习的框架进行反迭代投影的全局优化改进,改善了图像的重建质量。     

基于双稀疏模型的压缩感知核磁共振图像重构

医学核磁共振图像重构技术是核磁共振成像领域的关键技术之一。压缩感知理论指出利用核磁共振图像的稀疏性能够从高度欠采样的观测值中精确重构图像。如何利用图像的稀疏性先验以及更多的先验知识来提高重构质量成为核磁共振成像的一个关键问题。本文根据综合稀疏模型和稀疏变换模型的相互补充作用,利用核磁共振图像在这两种模型下的稀疏性先验,将结合了综合稀疏模型与稀疏变换模型的双稀疏模型应用于压缩感知核磁共振图像的重构系统,提出了一种融合双字典学习的自适应图像重构模型。本文充分利用了图像在自适应综合字典学习和自适应变换字典学习下的两种稀疏先验知识,使用交替迭代最小化法对提出的模型进行分阶段求解,求解过程中引入了综合K-奇异值分解(K-SVD)算法和变换K-SVD算法。通过实验验证,与目前较好的核磁共振图像重构模型对比,本文提出模型的图像重构效果更好、收敛速度更快,且具有更好的鲁棒性。     

Medical Image Fusion Based on Sparse Representation with KSVD

Medical image fusion is a process by which two different models of images are combined into a single image, in order to provide doctors with accurate diagnoses, and take right action. This paper proposes an image fusion method based on sparse representation with KSVD. Firstly, all source images are combined into a joint-matrix, which can be represented with sparse coefficients using an overcompletedictionary trained by KSVD algorithm. Secondly, the coefficients which are considered as image features are combined with the choose-max fusion rule. Finally, the fused image is reconstructed from the concatenated coefficients and the overcomplete dictionary. Compared with three state-of-the-art algorithms, the proposed method has better fusion performance.     

Feature-Motivated Simplified Adaptive PCNN-Based Medical Image Fusion Algorithm in NSST Domain

Multimodality medical image fusion plays a vital role in diagnosis, treatment planning, and follow-up studies of various diseases. It provides a composite image containing critical information of source images required for better localization and definition of different organs and lesions. In the state-of-the-art image fusion methods based on nonsubsampled shearlet transform (NSST) and pulse-coupled neural network (PCNN), authors have used normalized coefficient value to motivate the PCNN-processing both low-frequency (LF) and high-frequency (HF) sub-bands. This makes the fused image blurred and decreases its contrast. The main objective of this work is to design an image fusion method that gives the fused image with better contrast, more detail information, and suitable for clinical use. We propose a novel image fusion method utilizing feature-motivated adaptive PCNN in NSST domain for fusion of anatomical images. The basic PCNN model is simplified, and adaptive-linking strength is used. Different features are used to motivate the PCNN-processing LF and HF sub-bands. The proposed method is extended for fusion of functional image with an anatomical image in improved nonlinear intensity hue and saturation (INIHS) color model. Extensive fusion experiments have been performed on CT-MRI and SPECT-MRI datasets. Visual and quantitative analysis of experimental results proved that the proposed method provides satisfactory fusion outcome compared to other image fusion methods.     

An Improved Medical Image Fusion Algorithm for Anatomical and Functional Medical Images

In recent years,many medical image fusion methods had been exploited to derive useful information from multimodality medical image data, but, not an appropriate fusion algorithm for anatomical and functional medical images. In this paper, the traditional method of wavelet fusion is improved and a new fusion algorithm of anatomical and functional medical images,in which high-frequency and low-frequency coefficients are studied respectively. When choosing high-frequency coefficients, the global gradient of each subimage is calculated to realize adaptive fusion,so that the fused image can reserve the functional information;while choosing the low coefficients is based on the analysis of the neighborbood region energy, so that the fused image can reserve the anatomical image＇ s edge and texture feature. Experimental results and the quality evaluation parameters show that the improved fusion algorithm can enhance the edge and texture feature and retain the function information and anatomical information effectively.     

结合二维灰度直方图与PCNN的医学图像多类分割

针对医学图像中对组织器官多类分割的要求,提出一种结合二维灰度直方图的脉冲耦合神经网络（pulse coupled neural networks,PCNN）图像多类分割算法.首先根据PCNN模型的局部连接作用和阈值衰减特性对图像进行多类分割,然后利用基于类内最小离散度的二维直方图算法计算出PCNN网络迭代时的最佳门限值,从而实现医学图像的多类分割.通过对仿真的正常颅脑和非正常的颅脑核磁共振图像进行测试,结果显示本PCNN图像多类分割算法能够有效地分割出核磁共振颅脑图像中不同脑组织.因此,本文算法具有应用于医学图像的多类分割的可行性,并提高计算机辅助分割医学图像的准确性.     

Detail-enhanced multimodality medical image fusion based on gradient minimization smoothing filter and shearing filter

In this paper, a detail-enhanced multimodality medical image fusion algorithm is proposed by using proposed multi-scale joint decomposition framework (MJDF) and shearing filter (SF). The MJDF constructed with gradient minimization smoothing filter (GMSF) and Gaussian low-pass filter (GLF) is used to decompose source images into low-pass layers, edge layers, and detail layers at multiple scales. In order to highlight the detail information in the fused image, the edge layer and the detail layer in each scale are weighted combined into a detail-enhanced layer. As directional filter is effective in capturing salient information, so SF is applied to the detail-enhanced layer to extract geometrical features and obtain directional coefficients. Visual saliency map-based fusion rule is designed for fusing low-pass layers, and the sum of standard deviation is used as activity level measurement for directional coefficients fusion. The final fusion result is obtained by synthesizing the fused low-pass layers and directional coefficients. Experimental results show that the proposed method with shift-invariance, directional selectivity, and detail-enhanced property is efficient in preserving and enhancing detail information of multimodality medical images.

Open image in new window

Graphical abstract The detailed implementation of the proposed medical image fusion algorithm.

     

MRI-PET image fusion based on NSCT transform using local energy and local variance fusion rules

Image fusion means to integrate information from one image to another image. Medical images according to the nature of the images are divided into structural (such as CT and MRI) and functional (such as SPECT, PET). This article fused MRI and PET images and the purpose is adding structural information from MRI to functional information of PET images. The images decomposed with Nonsubsampled Contourlet Transform and then two images were fused with applying fusion rules. The coefficients of the low frequency band are combined by a maximal energy rule and coefficients of the high frequency bands are combined by a maximal variance rule. Finally, visual and quantitative criteria were used to evaluate the fusion result. In visual evaluation the opinion of two radiologists was used and in quantitative evaluation the proposed fusion method was compared with six existing methods and used criteria were entropy, mutual information, discrepancy and overall performance.