NMF-SVM based CAD tool applied to functional brain images for the diagnosis of Alzheimer's disease

This paper presents a novel computer-aided diagnosis (CAD) technique for the early diagnosis of the Alzheimer's disease (AD) based on nonnegative matrix factorization (NMF) and support vector machines (SVM) with bounds of confidence. The CAD tool is designed for the study and classification of functional brain images. For this purpose, two different brain image databases are selected: a single photon emission computed tomography (SPECT) database and positron emission tomography (PET) images, both of them containing data for both Alzheimer's disease (AD) patients and healthy controls as a reference. These databases are analyzed by applying the Fisher discriminant ratio (FDR) and nonnegative matrix factorization (NMF) for feature selection and extraction of the most relevant features. The resulting NMF-transformed sets of data, which contain a reduced number of features, are classified by means of a SVM-based classifier with bounds of confidence for decision. The proposed NMF-SVM method yields up to 91% classification accuracy with high sensitivity and specificity rates (upper than 90%). This NMF-SVM CAD tool becomes an accurate method for SPECT and PET AD image classification.     

基于流形学习的遮挡条件下表情识别方法

提出了一种识别遮挡图像表情的方法。先用主元分析(PCA,principal component analysis)算法对遮挡图像重建;然后根据正态分布理论检测出遮挡区域,并根据图像的部分相似性,将遮挡图像嵌入到流形空间中;最后用支持向量机(SVM,support vector machine)实现表情分类。本方法较好地消除了遮挡区检测误差对表情识别的影响,对遮挡图像的表情识别具有良好的鲁棒性。通过Cohn-Kanade和JAFFE人脸库上的表情识别实验,验证了本方法有较强的鲁棒性、较高的识别率和较高的运行效率。     

基于海马纹理特征的阿尔茨海默病早期识别

阿尔茨海默病(AD)是一种神经退行性疾病。随着脑医学影像的发展,对AD诊断的精确度也在进一步提高,但对AD的诊断,客观上仍缺少好的生物标记。为寻找到AD的更稳定的生物标记,利用海马的影像组学特征对海马的信号强度、形状、灰度阶梯分布等特征进行刻画,通过方差分析(ANOVA)和事后检验,在统计学上寻找出正常对照(NC)、AD、轻度认知损害(MCI)之间存在差异的特征;通过与被试的简易智能状况检查(MMSE)评分进行相关性分析,找寻与MMSE评分相关性较高的特征;利用支持向量机(SVM)构建一个对AD和NC分类的模型,交叉验证得到的正确率为86%。结果表明,海马的影像组学特征是一个很好的生物标记,能对AD进行有效的早期识别。     

An accurate traffic classification model based on support vector machines

Network traffic classification is a fundamental research topic on high‐performance network protocol design and network operation management. Compared with other state‐of‐the‐art studies done on the network traffic classification, machine learning (ML) methods are more flexible and intelligent, which can automatically search for and describe useful structural patterns in a supplied traffic dataset. As a typical ML method, support vector machines (SVMs) based on statistical theory has high classification accuracy and stability. However, the performance of SVM classifier can be severely affected by the data scale, feature dimension, and parameters of the classifier. In this paper, a real‐time accurate SVM training model named SPP‐SVM is proposed. An SPP‐SVM is deducted from the scaling dataset and employs principal component analysis (PCA) to extract data features and verify its relevant traffic features obtained from PCA. By employing PCA algorithm to do the dimension extraction, SPP‐SVM confirms the critical component features, reduces the redundancy among them, and lowers the original feature dimension so as to reduce the over fitting and increase its generalization effectively. The optimal working parameters of kernel function used in SPP‐SVM are derived automatically from improved particle swarm optimization algorithm, which will optimize the global solution and make its inertia weight coefficient adaptive without searching for the parameters in a wide range, traversing all the parameter points in the grid and adjusting steps gradually. The performance of its two‐ and multi‐class classifiers is proved over 2 sets of traffic traces, coming from different topological points on the Internet. Experiments show that the SPP‐SVM's two‐ and multi‐class classifiers are superior to the typical supervised ML algorithms and performs significantly better than traditional SVM in classification accuracy, dimension, and elapsed time.     

结合WGAN GP与CNN SVM的滚动轴承故障红外诊断

针对实际工程应用中由于滚动轴承故障状态出现的时间很短而导致数据集不平衡难以采用深度学习算法进行故障诊断的问题,提出了一种基于Wasserstein距离的梯度惩罚生成对抗网络(WGAN GP)和基于支持向量机分类的卷积神经网络(CNN SVM)相结合的滚动轴承故障红外诊断方法。从红外热像图中构建不平衡数据集,通过采用WGAN GP对不平衡数据扩充以达到数据集均衡,之后将CNN SVM模型应用于数据集,提取样本深度特征完成故障分类。实验表明,WGAN GP与CNN SVM相结合的模型在不平衡数据集下表现良好,相较于其他模型有更好的故障诊断能力,并且在故障分类阶段的用时可减少1689以上。     

基于Gabor滤波器和ICA技术的SAR目标识别算法

提出了一种基于Gabor滤波器和独立分量分析(ICA)技术对合成孔径雷达(SAR)目标识别的算法.该方法提取预处理后SAR图像的低频子带图像,利用Gabor滤波器组对该低频子带图像在不同方向和尺度上滤波,再用主成分分析(PCA)+ICA方法对Gabor滤波后图像提取有效特征向量作为目标识别特征,最后用支持向量机(SVM)对该特征进行分类完成目标识别.使用MSTAR数据库中3类SAR目标数据对该方法进行目标识别的仿真实验,平均识别率最高可达96.56%.通过与其他识别方法对比实验,验证了文中方法的有效性.     

Regions of interest extraction from SPECT images for neural degeneration assessment using multimodality image fusion

The aging population highlights the importance of early diagnosis of neurodegenerative diseases in the elderly. Current diagnoses of such diseases rely on visual assessment of the neuron activity of the specific regions in the brain revealed by SPECT imaging with a specific tracer, 99mTc-TRODAT-1. However, due to the difficulties in defining the regions of interest (ROI) in SPECT images, efficient indices are lacking for quantitative analysis. In this study, we performed simultaneous CT and SPECT scans and used the CT images as the medium to register the MR and SPECT images, such that the ROI delineated in the MR image can be mapped onto the SPECT image in the corresponding area. A robust registration scheme is proposed, including coarse registration using principal axes alignment and then fine-tuning the registration using a combination of maximal cross-section area detection and the general Hough transform. The results from three clinical datasets all show improved accuracy of registration as compared with the results obtained using conventional principal axes alignment alone. Based on these registration results, a correct ROI can be defined in the SPECT images and ROI-based quantitative indices can be further derived.     

Asymmetric Semi‐Supervised Boosting Scheme for Interactive Image Retrieval

Support vector machine (SVM) active learning plays a key role in the interactive content‐based image retrieval (CBIR) community. However, the regular SVM active learning is challenged by what we call “the small example problem” and “the asymmetric distribution problem.” This paper attempts to integrate the merits of semi‐supervised learning, ensemble learning, and active learning into the interactive CBIR. Concretely, unlabeled images are exploited to facilitate boosting by helping augment the diversity among base SVM classifiers, and then the learned ensemble model is used to identify the most informative images for active learning. In particular, a bias‐weighting mechanism is developed to guide the ensemble model to pay more attention on positive images than negative images. Experiments on 5000 Corel images show that the proposed method yields better retrieval performance by an amount of 0.16 in mean average precision compared to regular SVM active learning, which is more effective than some existing improved variants of SVM active learning.     

去噪声的加权SVM分类方法

针对支持向量机(SVM)本身抗噪声能力低和训练数据类别不均匀会造成分类结果偏向数目较大一类的倾向性等问题，本文提出了去噪声的加权SVM分类方法。在该方法中，通过引入主成分分析方法来降维去除噪声，再通过引入加权系数的方式，补偿了上述倾向性造成的不利影响，提高了预测分类精度。对污水处理过程运行状态的分类实验表明该方法的有效性。     

A linear correction for principal component analysis of dynamic fluorescence diffuse optical tomography images

The analysis of dynamic fluorescence diffuse optical tomography (D-FDOT) is important both for drug delivery research and for medical diagnosis and treatment. The low spatial resolution and complex kinetics, however, limit the ability of FDOT in resolving drug distributions within small animals. Principal component analysis (PCA) provides the capability of detecting and visualizing functional structures with different kinetic patterns from D-FDOT images. A particular challenge in using PCA is to reduce the level of noise in D-FDOT images. This is particularly relevant in drug study, where the time-varying fluorophore concentration (drug concentration) will result in the reconstructed images containing more noise and, therefore, affect the performance of PCA. In this paper, a new linear corrected method is proposed for modeling these time-varying fluorescence measurements before performing PCA. To evaluate the performance of the new method in resolving drug biodistribution, the metabolic processes of indocyanine green within mouse is dynamically simulated and used as the input data of PCA. Simulation results suggest that the principal component (PC) images generated using the new method improve SNR and discrimination capability, compared to the PC images generated using the uncorrected D-FDOT images.     

基于正交指数局部保留投影的高光谱图像特征提取

针对高光谱图像,在判别局部保留投影(Discri minant Locality Preserving Projection,DLPP)的基 础上,提出了一种名为正交指数判别局部保留投影(Orthogonal Exponential Discriminan t Locality Preserving Projection,OEDLPP)的特征提取方法。该算法不但保留了DLPP算法的有监督特性,还利用 了指数矩阵(the matrix exponential)来获取更有效的样本信息,避免了小样本问题。同时,OEDLPP 对投影矩阵进行 施密特正交化,解决了特征的冗余性问题。应用OEDLPP算法对高光谱图像进行特征提取后, 并采用支持 向量机(SVM)对降维后的数据进行分类。与主成分分析(PCA)、局部保留投影(LPP)、 判别局部保 留投影(DLPP)、指数判别局部保留投影(EDLPP)、正交判别局部保留投影(ODLPP)等对 比实验结 果表明,本文算法对样本有效信息的获取具有一定的优越性,分类精度提升了2%~3%左右。     

Automatic image orientation detection

We present an algorithm for automatic image orientation estimation using a Bayesian learning framework. We demonstrate that a small codebook (the optimal size of codebook is selected using a modified MDL criterion) extracted from a learning vector quantizer (LVQ) can be used to estimate the class-conditional densities of the observed features needed for the Bayesian methodology. We further show how principal component analysis (PCA) and linear discriminant analysis (LDA) can be used as a feature extraction mechanism to remove redundancies in the high-dimensional feature vectors used for classification. The proposed method is compared with four different commonly used classifiers, namely k-nearest neighbor, support vector machine (SVM), a mixture of Gaussians, and hierarchical discriminating regression (HDR) tree. Experiments on a database of 16 344 images have shown that our proposed algorithm achieves an accuracy of approximately 98% on the training set and over 97% on an independent test set. A slight improvement in classification accuracy is achieved by employing classifier combination techniques.     

Restoration of gamma camera-based nuclear medicine images

Experiments were conducted using a Siemens Rota camera to study the applicability of two linear shift-invariant (LSI) filters, namely, the Wiener and power spectrum equalization filters, for restoration of planar projections and single-photon-emission computed tomography (SPECT) images. In the restoration scheme, the system transfer function, computed from a line source image, is modeled by a 2-D Gaussian function. The noise power spectrum is modeled as a constant for planar images and as a ramp for SPECT images. The filters have been applied to restore computer-simulated 1-D and 2-D projections and SPECT images of two simple phantoms, 2-D projections of two phantoms obtained from the Siemens Rota camera, and SPECT images of a cardiac phantom obtained from the Siemens Rota camera. The filters are shown to perform partial restoration. Considerable noise suppression and detail enhancement have been observed in the restored images. quantitative measurements such as root-mean-squared error and contrast ratio have been used for objective analysis of the results, which are encouraging.     

Quantitative rotating multisegment slant-hole SPECT mammography with attenuation and collimator-detector response compensation

Rotating multisegment slant-hole (RMSSH) single photon emission computed tomography (SPECT) is suitable for detecting small and low-contrast breast lesions since it has much higher detection efficiency than conventional SPECT with a parallel-hole collimator and can image the breast at a closer distance. Our RMSSH SPECT reconstruction extends a previous rotation-shear transformation-based method to include nonuniform attenuation and collimator-detector response (CDR) compensation. To evaluate our reconstruction method, we performed two phantom simulation studies with 1) an isolated breast and 2) a breast phantom attached to the body torso. The reconstructed RMSSH SPECT images with attenuation and CDR compensation showed improved quantitative accuracy and less image artifacts than without. To evaluate the clinical efficacy of RMSSH SPECT mammography, we used a simulation study to compare with planar scintimammography in terms of the signal-to-noise ratio (SNR) value of a breast lesion. The RMSSH SPECT reconstruction images showed higher SNR value than the planar scintimammography images and even more so as we applied compensation for attenuation and collimator detector response. We conclude that attenuation and CDR compensation provide RMSSH SPECT mammography images with improved quality and quantitative accuracy.     

基于主成分分析和支持向量机的飞参阶段划分研究

主成分分析是一种应用广泛的线性降维技术,它在保留住数据的重要成分的同时达到了对数据的降维。对高维、多属性的飞参数据进行主成分分析,可以实现飞参的降维。支持向量机的学习方法则以其全局最优和泛化能力好的特．最,实现对飞参阶段的划分。使用主成分分析后的数据进行阶段划分可以提高划分速度,并且划分效果更好。     

最佳PCA稀疏表示方法及在直升机旋翼故障识别中的应用

提出一种基于PCA稀疏表示算法进行直升机旋翼故障识别的方法。首先相应的PCA预处理本身计算复杂度不高且能对样本的降维程度较高,其次根据样本相似性原则,基于PCA的稀疏表示方法不仅能保持样本在处理前后相互距离不变,而且提高了计算效率。采用新的诊断模型对直升机旋翼故障分类识别,并与基于神经网络和基于支持向量机的诊断方法进行比较。结果表明本文方法对旋翼故障具有良好的识别能力。     

Two-dimensional restoration of single photon emission computed tomography images using the Kalman filter

The discrete filtered backprojection (DFBP) algorithm used for the reconstruction of single photon emission computed tomography (SPECT) images affects image quality because of the operations of filtering and discretization. The discretization of the filtered backprojection process can cause the modulation transfer function (MTF) of the SPECT imaging system to be anisotropic and nonstationary, especially near the edges of the camera's field of view. The use of shift-invariant restoration techniques fails to restore large images because these techniques do not account for such variations in the MTF. This study presents the application of a two-dimensional (2D) shift-variant Kalman filter for post-reconstruction restoration of SPECT slices. This filter was applied to SPECT images of a hollow cylinder phantom; a resolution phantom; and a large, truncated cone phantom containing two types of cold spots, a sphere, and a triangular prism. The images were acquired on an ADAC GENESYS camera. A comparison was performed between results obtained by the Kalman filter and those obtained by shift-invariant filters. Quantitative analysis of the restored images performed through measurement of root mean squared errors shows a considerable reduction in error of Kalman-filtered images over images restored using shift-invariant methods.     

基于生物免疫优化支持向量机算法的居民区负荷预测

针对居民区用电负荷随机性强、稳定性差等问题,综合考虑各因素对居民用电负荷的影响,提出一种免疫支持向量机(support vector machine,SVM)算法负荷预测模型。以居民区历史用电量及相关气候数据为处理对象,使用PCA(principal component analysis)算法对电网历史数据进行处理,并结合免疫算法对电网历史数据进行预处理,形成数据簇并划定标签提供给预测模型进行训练。为提高模型精度,采用生物免疫优化算法对SVM模型参数进行优化,并在负荷预测环节,将预测误差作为调优依据,对预测模型进行反馈调优。将预测效果与常用于负荷预测的BP(back propagation)神经网络、SVM算法模型进行对比,免疫SVM算法负荷预测模型的短期、中期预测精准度均在98%以上,具有较好的精度与鲁棒性。     

Aperture collimation correction and maximum-likelihood image reconstruction for near-field coded aperture imaging of single photon emission computerized tomography

Coded aperture (CA) imaging originally developed in X-ray astronomy has not been widely used in nuclear medicine due to the decoding complexity of near-field CA images. In this paper, we present a near-field CA imaging technique and image reconstruction method for high sensitivity and high resolution single photon emission computerized tomography (SPECT). Our approach makes three contributions. First, a correction method for the aperture collimation effect is used to eliminate the near-field artifacts without dual acquisitions of mask and anti-mask images. Second, a maximum-likelihood expectation-maximization (MLEM) deconvolution method is used to restore CA images. Finally, a new MLEM-based algorithm is used to partially reconstruct three-dimensional (3-D) objects from a single projection of CA images. Experiments conducted using a dual-head SPECT system equipped with a parallel-hole collimator and a CA module show a tenfold increase in count sensitivity and significant improvement in image resolution with CA collimation as compared to parallel-hole collimation. Experiments conducted using the same dual-head SPECT system equipped with a pinhole collimator show that when the object is closer to the pinhole collimator the CA image resolution is only slightly inferior to the pinhole collimated image. We found that the MLEM deconvolution method provides an inherent nonnegativity constraint on pixel values and remarkably reduces background activities of CA images. The MLEM reconstruction algorithm for CA images is capable of reconstructing 3-D objects from a single projection and can be potentially extended to full 3-D SPECT reconstruction for CA images.     

电力设备IR图像特征提取及故障诊断方法研究

针对电力设备红外图像批量诊断中故障特征参量提取及参数配置难题,采用粒子群算法(PSO)与Niblack算法相结合的方法,将设备热像从背景中分割出来并提取出设备的最低、最高及平均温度等参量,通过计算设备各温升特征,构建支持向量机(SVM)样本特征空间。采用优化的蝙蝠算法(BA)对SVM参数进行寻优,并利用最优参数配置下的SVM实现设备故障诊断。对220组图像样本测试结果表明：该红外图像故障诊断方法在电力设备热故障缺陷检测方面的效率及准确率较高,适用于电力大数据中非结构化红外图像的批量分析与处理。