Introduction

Proteomics is the analysis of the complete set of proteins(i.e.,the proteome)produced in a cell,tissue,or organism,at a given time.A number of different aspects of protein analysis are covered in proteomics,including the analysis of protein expression.The presence or absence of proteins,and the direct measurement of relative protein abundances can help understand cellular processes,and may be useful for identifying drug targets and diagnostic/prognostic markers[1].As such,it is one of the most active fields of biological research,given its wide range of applications.

废水中有机物的降解方法需要更加深入的实验研究，寻找降解效果更好的氧化剂，结合选矿废水处理的实际情况，提高降解效率将是后一步实验方向。

第二，事业单位管理观念陈旧。事业单位的管理制度与政府部门的管理制度是一致的，很多制度内容对于目前的市场环境来说，显然已经落后了。若是不能尽早发现问题，解决管理观念上存在的问题，将会严重影响事业单位的工作质量，严重影响事业单位工作指标的完成。

One of the most commonly used techniques in comparative proteomic studies is two-dimensional gel electrophoresis(2-DGE)[2],a technique capable of resolving thousands of proteins in a single run.Using 2-DGE,proteins are firstly separated according to their isoelectric points(pIs)in one dimension,and then,according to their molecular weights in a second dimension[3].Afterward,the 2-DGE gels are stained for protein visualization and analyzed with computer-assisted image evaluation systems for a comprehensive qualitative and quantitative examination of the proteomes[4].

另一方面，树立财务职业素养。大数据时代下虽然为财务管理提供了多样化的资源，但资料的数据化和信息化使得财务管理工作以及企业运作的不安全性因素增加，如企业财务数据被盗或者低素质人员基于个人私利出卖企业数据等，因此，有必要引导财务人员树立专业化的素养，保证数据和信息安全的前提下开展财务管理工作。

2-DGE was initially employed for protein separation and analysis approximately in 1975[5].Even though it was introduced more than four decades ago,nowadays 2-DGE is still widely used for whole proteome analysis[6,7],comparative analysis of proteome changes[6,8],biomarker discovery,cancer research[9,10],as well as for the identification of protein isoforms and post-translational modifications[11],among other purposes.Its popularity could be attributed to several factors.It is possible to resolve and visualize thousands of proteins in a single gel as mentioned above.In addition,2DGE is very affordable compared to other techniques,which does not require complex laboratory equipment but offers reliable and reproducible results.2GDE is optimal for cases when multiple analyses are required on replicate samples.More importantly,it is compatible with mass spectrometry and other downstream analyses.Therefore,2GDE is often used as one of the first steps in describing the protein composition of a particular sample type,at a certain time point,and under a particular set of conditions.

Many computational applications are available for processing and analyzing 2-DGE images,such as MELANIE,PDQuest,Z3,Progenesis Workstation,ProteomeWeaver,ProteinMine,Delta2D,and DeCyder[12,13].Given that one cell can express around ten thousand proteins,2-DGE needs an effective computational tool that can process large volumes of information[14].It is often necessary to apply appropriate preprocessing techniques to 2-DGE images,as the ultimate performance of these analysis tools strongly depends on the quality of the images to be processed[4,15].

For 2-DGE image analysis,techniques are required to detect protein spots,to segment and to quantify the protein expression level based on the number of pixels[15,16].An additional step of image alignment is performed in order to match the corresponding protein spots from different images[17].However,due to technical difficulties inherent to 2-DGE,anomalies are often found in gel images,such as noise around protein spots,vertical and horizontal streaking,saturation of certain protein spots,presence of very faint protein spots,as well as non-linear intensity of protein spots[14,18].Therefore,2-DGE image analysis is frequently perturbed by the presence of different types and levels of noise.For instance,the impulsive noise can spectrally interfere in all frequencies when the sample is digitalized.The background of 2-DGE images can also vary among samples,depending on the technical specifications of the imaging system used to capture and digitalize the images[15].These variations and anomalies often complicate the analysis of 2-DGE images and affect the reproducibility of the results obtained[19,20].A preprocessing phase could be effective in reducing or eliminating these anomalies,thus reducing errors in the subsequent spot detection.Omitting this preprocessing stage may profoundly affect the results of the subsequent analysis,as noise could be falsely identified as a protein,whereas real proteins of interest could be missed[21].Therefore,it is important to review the advances in 2-DGE image preprocessing,in order to identify areas of improvement that may be of great interest to researchers in the field.

There are three common objectives for the 2-DGE image preprocessing,i.e.,pixel intensity regularization(image normalization),background correction,and noise reduction(filtration)[15].In the subsequent sections,we describe in detail the preprocessing strategies to achieve these objectives,their limitations,and recent advances.We also compare the most representative noise reduction techniques using synthetic gel images for a quantitative evaluation.Finally,we conclude with recommendations to be considered for successful automatic 2-DGE image processing.

Preprocessing of 2-DGE images

2-DGE images inevitably exhibit anomalies caused by sample preparation techniques and the imaging system used to acquire the digital images[22].The most common anomalies in 2-DGE images are oversaturated,faint,or fuzzy spots,vertical and horizontal streaking,overlapping spots,and noise[13,23].Figure 1 shows a 2-DGE image with the most common anomalies.The image was obtained in our laboratory from prostate tissue,the challenges of image preprocessing,and thus,the requirement for robust algorithms in order to achieve proper preprocessing are apparent.

Image normalization

Orthogonal regression methods have been used as a simple and ef ficient option that can adapt to a linear model[41].However,the edges of protein spots of interest in 2-DGE images are often distorted during filtering[25,42].The linear filters based on orthogonal regression techniques,which are most commonly used for noise reduction in 2-DGE images,are based on principal component analysis(PCA)and Wiener filters.PCA is a robust orthogonal regression technique,which shows a great advantage in terms of spot edge maintenance.However,small details,such as low-abundance protein spots,can be removed in the process[43,44].

To avoid saturation caused by high protein concentrations,it is necessary to make sure that most of the data are represented by levels slightly below the maximum possible intensity.Highly abundant proteins often mask fainter spots that represent proteins with low abundance.Usually,these lowabundance proteins are also relevant,but pose difficulties in accurate quantification,if the spots are too faint[27].

The normalization process helps to reduce noise,which distorts the detection of protein spots in 2-DGE images according to criteria of statistical significance,thus allowing the outline of the spots of interest to be delineated[28].However,in some circumstances,the intensity of pixels does not present a clear tendency.This type of standardization is based on information from various replicate gels,generating a statistical average of the intensity of each pixel.With these statistical averages,the intensities of the entire image are normalized,using normalization algorithms such as Cyclic Loess,Contrast,Cyclic Linear,and Quantile[28].Image regularization with multiple replicate gels has yielded satisfactory results,since changes in patterns of protein expression among samples are small compared to the overall pattern of protein spots.The major limitation of this approach lies in obtaining and aligning a representative number of gel images,which may result in the unwanted removal of faint protein spots[15].With image regularization,it is possible to minimize distortions present in the images;however,low-abundance proteins are often eliminated in the process.Therefore,it is necessary to apply discriminatory algorithms to these regularization techniques,so that potentially faint but important protein spots would not be removed during image regularization.

Background correction

The aim of 2-DGE image processing is to detect spots,and define the location,border,and intensity of each spot.However,in some conditions,defining the limits of the spot border is difficult due to the varied background,noise,and saturation of protein spots[14,29,30].Background correction is thus necessary,as the background signal of 2-DGE images is not uniform and varies locally according to protein intensities.The greater the intensity of a protein spot,the greater the noise signal around it,which introduces errors in detection.In addition,anomalies such as vertical streaking must be removed from the image,as protein spots contained within these streaks are not identifiable or quantifiable[15,31].Three background correction techniques are available to reduce the influences of these phenomena.These include adjustment by polynomials,local and global minima,and histograms.

2-DGE image filtering is the most commonly used approach within the scientific community,with the most diversified techniques[15].These filtering techniques can be classified into two broad categories:linear and nonlinear filters.

For background adjustment using histograms,the distribution of intensities in an image is used to identify the background.The maximum peak in the intensity distribution is the intensity of background pixels.The challenge of this technique is to establish the intensity threshold of the background that is usually set by an optimization process[34].The selection of parameters is a signi ficant drawback for users with little knowledge of optimization techniques,considering that the model implemented to describe the background may be insufficient or cause a signi ficant loss of faint but important protein spots[35].Although studies on the implementation of background correction techniques for 2-DGE images have lost momentum,a variety of techniques[36]in this field are yet to be explored.In addition,new techniques have also been developed that are efficient for the analysis of images with features similar to those of 2-DGE images,such as combination of Gaussian mixture models and machine learning approach[37],which may be promising options for 2-DGE image processing.

Noise reduction

Background correction with polynomial adjustment has been widely accepted in the scientific community[32].This technique seeks smoothing the image background,removing noise from overlapping spots,and minimizing the effects of intensity degradation.The biggest limitation of background correction is the selection of a background sample for the adjustment.A poor selection of the background sample can generate significant image distortion,or even lead to the improper softening of background texture[33].Local and global minima are the simplest technique for background correction,making it possible to effectively mitigate noise in the flatter areas(without signi ficant changes in pixels)and smooth discontinuities.However,a global threshold is not suf ficient to detect background variations[15].

（2）注浆材料。注浆材料宜采用水泥浆，水泥必须采用普通硅酸盐水泥，标号42.5，可添加适量添加剂，水和水泥配比应不大于0.5，水泥浆密度不应小于1.90 g/cm3。

当然，在国家权力和政策调试下，适应政府治理需求只是网络政治参与娱乐化产生的一个方面。相反，个体网络行为的选择首先是要满足自我政治参与的需求，使得网络政治参与成为个体表达诉求的有效渠道。而网络政治参与娱乐化的灵活性与抗争性优点，无疑为个体表达政治诉求提供了有效手段。

The first techniques reported for 2-DGE image’preprocessing were based on linear filters,which are the most widely used techniques in commercial software programs.These kinds of filters compute the output pixel as a linear relation of the surrounding pixels[38].In general,linear filters tend to blur images,affecting the edges,lines,and fine details of the protein spots[39].While linear filtering techniques,such as the Gaussian filters,have an acceptable performance under general conditions,they are also likely to reduce the density of proteins spots by smoothing the contour[40].

Intensity normalization of 2-DGE images aims to normalize the intensity of protein spots to facilitate processing.The spot intensities may vary due to differences in the exposure of the gel during the staining process and data saturation,among other reasons[15].As a step of image processing[24],intensity normalization is also known as a preprocessing technique in various literatures[25].In fact,some researchers recommend eliminating saturated spots before processing in order to achieve the optimal results for the identification of proteins in the gel[14,26].

The lack of performance indicators with which the various techniques can be fairly compared greatly hinders the selection of preprocessing techniques.In fact,performance comparison of many commonly used techniques has not yet been reported in the literature[15].The most commonly used indicator is the signal-to-noise ratio(SNR)obtained from the mean square error(MSE):

Nonlinear filters have fewer limitations,but generally increase the degree of complexity and the need to define parameters more thoroughly.Some nonlinear filtering techniques used in 2-DGE images include contourlet transform[46],total variation[47],and wavelet transform[19].The contourlet transform performs a decomposition of the data in frequency and space by means of two decomposition methods[46].These methods allow a multiple scale of space and frequency,as well as a high degree of directionality.Satisfactory results have been achieved using these methods,in addressing the low signal-to-noise ratio[48],and loss of relevant information,thereby surpassing the wavelet transform[48].

The total variation technique,proposed by Chan[49]and Xin[50],is very advantageous in terms of the extraction of geometric features and in terms of the preservation of edges.Using this technique,different degrees of smoothing on different points can be achieved,which exemplifies its anisotropy.Although able to capture all the directional information,this technique leads to staircase effects in the noise removal process[51].Therefore,this technique does not perform as well as the wavelet transform technique[41,50].Wavelet transform is the most commonly used technique for noise reduction,with a high percentage of filtering and acceptable level of loss of relevant information[19,43,52–57].This method has become increasingly popular in the preprocessing of 2-DGE images since 2004,as reported in the study by Kaczmarek and others[19].

Despite the simplicity ofnoise suppression methodsbased on spatial filtering,the image tends to be severely distorted after processing.As these types of images have local features with high variation,it is almost impossible to discriminate between information of interest and noise,when the processing considered only the spatial or spectral domain[41].For this reason,it is necessary to use a filter with a joint time–frequency domain[57]that exceeds spatial filtering,in both signal/noise ratio and distortion.Using the wavelet transform method,the image can be decomposed into space-frequency components and evaluated point by point.It is also possible to combine several images in order to obtain an average value and thus eliminate noise more effectively[56].For a correct operation,it is necessary to determine the scale of the wavelet transform.However,itis difficultto properly define the contours,resulting in a significant distortion of the information of interest[50].

In terms of spot efficiency,wavelet and WVTV techniques have a lower sensitivity to noise than the contourlet and TV techniques,as reflected by the best performance of wavelet and WTTV techniques.Wavelet was the best for images with noise levels between 12 and 20 dB,and WVTV was best for images with high-level noise.Figure 3 presents an example of filtering results obtained for the image with Gaussian noise equal to 10 dB.

However,with the SNR indicator,important factors regarding the retention of relevant information are not considered.Daszykowski thus proposed a measure called false discovery rate(FDR),defined as a ratio of the number of stable variables from the permuted data to the number of stable variables from the original dataset[25].While this measure is focused on the object detection process,it does not list the spots of interest that were removed during the preprocessing routine.FDR can be estimated by:

Validation measures for processing 2-DGE images

With the Wiener filter,it is possible to obtain satisfactory results with a low level of distortion of the image,but the noise reduction is compromised[38].When taking into account the overall advantages of noise removal,a certain level of signal degradation is generally acceptable.However,the noise reduction results in the blurred image with poor edge definition[15].In general,these methods allow the user to select the number of times the smoothing filter is applied,and the size of the convolution window to adjust the level of smoothing[45].

Due to the limitations of these techniques and the need for parameter adjustments,advanced techniques such as genetic algorithms[52]have been applied.However,it is noteworthy that this entails an increase in computational complexity.Although methods are available to minimize the computational cost of genetic algorithms,these add a higher level of complexity to the whole process[61].

商务合同是指有关各方之间在进行某种商务合作时，为了确定各自的权利和义务，而正式依法订立的、并且经过公证的、必须共同遵守的协议条文。(百度百科)作为两种语言的文化桥梁，商务合同是各国商人成功合作并维护双方权益的必经之路。为了保证国际经济活动的顺利进展和合同双方的权利，商务合同的翻译应该简明规范、语义清晰、符合汉语表达习惯。

where N1 is the number of falsely identified significant features in the permuted data for a given threshold value and N2 is the number of significant features in the experimental data set for the same threshold.

Another measure is spot efficiency[41]as defined below,which establishes a relationship between the false spots that were generated or prevailed,and spots that were lost with the technique.

Quantitative comparison of noise reduction methods

Comparison of synthetic images

The performance of several noise reduction techniques was compared using synthetic images.Experiments were developed under controlled conditions and the quantitative indicators SNR and spot efficiency were used to determine the performance of the filtering techniques with a perturbation of impulsive-type noise.To generate a cloud of spots,we used a model of uniform distribution,which simulates proteins in 2-D gels,and allowed overlapping between spots.The synthetic spot model is a Gaussian distribution[62],with adjustable parameters for size,intensity,and degradation(see Figure 2A).We used a 512×512 synthetic image with 250 protein spots.For each protein,the standard deviation was randomly set between 0.3 and 0.8,and its maximum intensity was constrained between 0.6 and 1.Different types of noise were added to the synthetic images to examine how the change of distribution affects the behavior of the techniques being evaluated.The noise distributions applied to the images are:Gaussian noise,Rayleigh type noise,and exponential type noise[63].Table 1 presents the parameters used for each noise type to simulate images with SNR between 8 and 20 dB.An example of a synthetic image with noise is presented in Figure 2B.

To obtain images with a blurring effect,a linear approximation of the camera movements in pixels(N)was performed,using image processing techniques with a filter of displacement vector[64].We used synthetic images with linear displacement from 5 to 30 pixels and a representative synthetic image with blurring effect is shown in Figure 2C.In addition,we also generated images with distortions between 5%and 30%in the background to represent a non-uniform illumination system,as exemplified by the image in Figure 2D.

The SNR and spot ef ficiency for the images with noise and filtered with nonlinear techniques wavelet,contourlet,TV,and WVTV are presented in Table 2 and Table 3,respectively.The technique with the best performance for each noise distribution,variance,and indicator is highlighted.As expected,the performance of each of these techniques decreases with the increasing noise level.In terms of SNR,wavelet is the best technique for images with Gaussian noise(Table 2).For images with Rayleigh and exponential noise,wavelet presents the best performance for low-level noise.Conversely,TV was better for high level of Rayleigh and exponential noise(8–12 dB).

对社会的管理是一项复杂的工程，需要公众对官员的高度信任。有人认为，官员是国家事务的管理者，凡为官者应当是社会的精英。近年来，政府官员的公信力频频受到质疑，引起人们的关注。官员要明确自身的社会责任，以民为本，真正做到为百姓谋福利，这样才能使国家保持和谐的状态。对此，在现实生活中，地方官员应学会为官之道，想民之所想，急民之所急。具体应做到一下几点：

Different derivations of the wavelet transform[58,59]are available to deal with its limitations.For instance,Barlaud and colleagues[60]presented a very useful technique called Wavelet Transform Quincunx,to obtain a multiresolution scale,thereby improving the performance around the edges and enhancing filtering effect[41].A hybrid technique,WTTV,combines the power of wavelets with the advantages of Total Variation(TV)[50],to reduce the loss of relevant information and increases the level of filtering.The wavelet transform can be applied first,followed by a total variation routine,thus minimizing distortions caused by the wavelet transform.In this case,it is necessary to carefully define the parameters for both components,since false information is generally introduced due to limitations of the directional information.As a result,the geometric captures can be affected,creating a distortion of the borders around the spots of interest[46].

Table 4 presents SNR and spot efficiency for images with the blurring effect and non-uniform background distribution filtered with nonlinear techniques wavelet,contourlet,TV,and WVTV.Contourlet and wavelet transform have the best performance with blurred images and images with a nonuniform background distribution.Both filtering techniques achieved comparable results for the different levels of blurring and distortion.

Comparison with a real 2-DGE image

A real 2-DGE image taken from the LECB 2-D PAGE Gel Images Datasets was used to compare the performance of the aforementioned filtering techniques[65].Sample 19,an annotated 2-DGE image of human leukemia(Figure 4),in which many proteins have been previously identified was selected for computing the rate of detection of these known proteins before and after filtering using the different techniques.Table 5 presents the rate of detected proteins from the original image,and of the image filtered by wavelet,contourlet,TV,and WVTV.It can be seen that wavelet,contourlet,and WVTV improved the rate of detection in comparison with the non-filtered image.A detection rate of 71.2%was achieved using TV,whereas a maximum detection rate of 87.9%was achieved using wavelet.

Remarks

The first step in most commercial 2-DGE image analysis software programs is spot detection.Often,false positive spots are detected due to noise and other artifacts in the image,and conversely,some real data may be missed.Therefore,image preprocessing is a fundamental step for proteomic analysis using 2-DGE,as the performance of the subsequent analysis is directly affected by the quality of the image.The development of image acquisition systems and optimal sample preparation protocols greatly improves image quality,reducing the intensity and frequency of unwanted anomalies.However,some anomalies are unavoidable and affect the accuracy and reproducibility of the proteomic characterization of the sample.Each of the preprocessing approaches meets particular needs.Nonetheless,the order these preprocessing stages should be implemented in,and whether all of them are actually necessary or some could be excluded without affecting the final performance,is still not clear.In conclusion,a methodology that adequately integrates these preprocessing approaches is needed.

在家政服务由传统服务业向现代服务业的转变过程中，红枫家政公司积极推进自身品牌建设，先后推出“红枫月嫂”“红枫育儿嫂”“红枫家务”“红枫保洁”“红枫培训”等一系列家政服务和家政培训项目，并积极参与“四川省妇联巾帼家政天府妹子”等品牌建活动，于2012年12月创建成为国家商务部认定的绵阳市家政服务行业龙头企业，于2013年2月成功注册“红枫”商标，于2015年12月被发展家庭服务业促进就业部际联席会议办公室评定为“全国千户百强家政服务企业”。

Even though several specialized image processing software programs are available for 2-DGE,these still require considerable human intervention,which affects reproducibility of the results as well.The wide range of anomalies in 2-DGE images,such as fuzzy spots,vertical and horizontal streaking,and noise,make it difficult to process the image.Therefore,it is important to implement advanced preprocessing techniques that could help mitigate the effect of these anomalies.In this review,we grouped three approaches of preprocessing:filtering,image regularization,and background correction,which are suited to different purposes such as noise reduction,regularization of the intensity of the pixels,and background correction.If these preprocessing procedures are not implemented,negative effects are observed during the image processing stage,such as image distortion,loss of relevant information,and changes in the contour of the spots of interest,yielding non-representative data.Table 6 summarizes the most common techniques,and the disadvantages of the three main groups of approaches for 2-DGE image pre-processing.

平日中默契十足，生活中是亲密无间的闺蜜，在养育宝宝方面更是相互借鉴彼此鼓励，她们俩就是吴瑶和coco。

Noise reduction is successful when filtering techniques are applied.Although linear filters are easy and simple for reducing noise in 2-DGE images,these filters tend to distort the edges of the spots of interest.Techniques such as Gaussian filters,orthogonal regression,and PCA often remove small details such as low-abundance proteins in the process.Thus,nonlinear noise reduction techniques are presented as moreappropriate solutions.Nonlinear filters are more robust,but also more complex to use;they require the definition of parameters that depend on the image.It is clear that there is still a great scope for improving the performance of preprocessing techniques for 2-DGE images,as well as for proposing new ones,since the decision to implement these techniques in 2-DGE image processing is based on their performance with other types of images.Therefore,2-DGE images are an attractive alternative for testing new algorithms for noise reduction,background correction,and image adjustment.

3) 解吸时间。已吸附花青素后的树脂，加入70%的乙醇溶液经25℃摇床振荡解吸1 h、2 h、4 h、8 h及12 h，过滤定容至25 mL，测其吸光度并计算解吸率。

In this article,we validated the quantitative performance of the most representative preprocessing techniques for noise reduction of 2-DGE images.By comparing the performance of wavelet,contourlet,TV,and WVTV techniques using SNR and spot efficiency indicators,we conclude that spot eff iciency is more appropriate for evaluating the noise reduction techniques applied to 2-DGE images.This index provides information about the effect of the noise reduction technique on the process of protein detection.Quantitative performance comparison using synthetic images indicates that wavelet filtering is the best technique under the test conditions,which achieves good results for the images with different types and levels of noise evaluated.

Taking into account the results obtained for the images with a blurring effect,and the fact that this anomaly is recurrent[66],we highlight the importance of high-quality image acquisition equipment,as this blurring effect is not mitigated effectively by the preprocessing techniques evaluated in this review.Blurring dramatically affects the performance of detection algorithms,and hence,specialized deblurring algorithms could be included in the preprocessing stage of 2-DGE images.However,cautions should be taken when including these techniques since they can lead to a significant distortion of the image[67].

Competing interests

The authors have declared no competing interests.

This work was supported by the Instituto Tecnologico Metropolitano(ITM)of Medellin,Colombia(Grant No.P14227)awarded to SR.

References

[1]Shruthi BS,Vinodhkumar Selvamani P.Proteomics:a new perspective for cancer.Adv Biomed Res 2016;5:67.

[2]Magdeldin S,Enany S,Yoshida Y,Xu B,Zhang Y,Zureena Z,et al.Basics and recent advances of two dimensional-polyacrylamide gel electrophoresis.Clin Proteomics 2014;11:16–26.

[3]Friedman DB,Hoving S,Westermeier R.Isoelectric focusing and two-dimensional gel electrophoresis. Methods Enzymol 2009;463:515–40.

[4]Rabilloud T,Lelong C.Two-dimensional gel electrophoresis in proteomics:a tutorial.J Proteomics 2011;74:1829–41.

[5]Klose J.Protein mapping by combined isoelectric focusing and electrophoresis of mouse tissues.A novel approach to testing for induced point mutations in mammals. Humangenetik 1975;26:231–43.

[6]Mitra A,Boroujeni MB.Application of gel-based proteomic technique in female reproductive investigations.J Hum Reprod Sci 2015;8:18–24.

[7]Curreem SOT,Watt RM,Lau SKP,Woo PCY.Two-dimensional gel electrophoresis in bacterial proteomics. Protein Cell 2012;3:346–63.

[8]Murphy S,Dowling P,Ohlendieck K.Comparative skeletal muscle proteomics using two-dimensional gel electrophoresis.Proteomes 2016;4:27.

[9]Huang CH,Chiou SH.Clinical proteomics identifies potential biomarkers in Helicobacter pylori for gastrointestinal diseases.World J Gastroenterol 2014;20:1529–36.

[10]alvarez-Chaver P,Otero-Este´vez O,de la Cadena MP,Rodri-guez-Berrocal FJ,Martinez-Zorzano VS.Proteomics for discovery of candidate colorectal cancer biomarkers.World J Gastroenterol 2014;20:3804–24.

[11]Rogowska-Wrzesinska A,Le Bihan MC,Thaysen-Andersen M,Roepstorff P.2D gels still have a niche in proteomics.J Proteomics 2013:4–13.

[12]Bandow JE,Baker JD,Berth M,Painter C,Sepulveda OJ,Clark KA,et al.Improved image analysis workflow for 2-D gels enables large-scale 2-D gel-based proteomics studies-COPD biomarker discovery study.Proteomics 2008;8:3030–41.

[13]Sengar RS,Upadhyay AK,Singh M,Gadre VM.Analysis of 2D-gel images for detection of protein spots using a novel nonseparable wavelet based method.Biomed Signal Process Control 2016;25:62–75.

[14]Shamekhi S,Miran Baygi MH,Azarian B,Gooya A.A novel multi-scale Hessian based spot enhancement filter for two dimensional gel electrophoresis images.Comput Biol Med 2015;66:154–69.

[15]Rye M,Fargestad EM.Preprocessing of electrophoretic images in 2-DE analysis.Chemom Intell Lab Syst 2012;117:70–9.

[16]Nhek S,Tessema B,Indahl UG,Martens H,Mosleth EF.2D electrophoresis image segmentation within a pixel-based framework.Chemom Intell Lab Syst 2015;141:33–46.

[17]Kostopoulou E,Katsigiannis S,Maroulis D.2D-gel spot detection and segmentation based on modified image-aware grow-cut and regional intensity information.Comput Methods Programs Biomed 2015;122:26–39.

[18]Smith R.Two-dimensional electrophoresis:an overview.Methods Mol Biol 2009;519:1–16.

[19]Kaczmarek K,Walczak B,De Jong S,Vandeginste BGM.Preprocessing of two-dimensional gel electrophoresis images.Proteomics 2004;4:2377–89.

[20]Rabilloud T,Chevallet M,Luche S,Lelong C.Two-dimensional gel electrophoresis in proteomics:past,present and future.J Proteomics 2010;73:2064–77.

[21]Tsakanikas P,Manolakos ES.Improving 2-DE gel image denoising using contourlets.Proteomics 2009;9:3877–88.

[22]Nhek S,Mosleth E,Høy M,Griessl M,Tessema B,Indahl U,et al.Nonlinear visualisation and pixel-based alignment of 2D electrophoresis images.Chemom Intell Lab Syst 2012;118:97–108.

[23]Savelonas MA,Mylona EA,Maroulis D.Unsupervised 2D gel electrophoresis image segmentation based on active contours.Pattern Recognit 2012;45:720–31.

[24]Chang J,Van Remmen H,Ward WF,Regnier FE,Richardson A,Cornell J.Processing of data generated by 2-dimensional gel electrophoresis for statistical analysis:missing data,normalization,and statistics.J Proteome Res 2004;3:1210–8.

[25]Daszykowski M,Wrobel MS,Bierczynska-Krzysik A,Silberring J,Lubec G,Walczak B.Automatic preprocessing of electrophoretic images.Chemom Intell Lab Syst 2009;97:132–40.

[26]Berth M,Moser FM,Kolbe M,Bernhardt J.The state of the art in the analysis of two-dimensional gel electrophoresis images.Appl Microbiol Biotechnol 2007;76:1223–43.

[27]Hortin GL,Sviridov D.The dynamic range problem in the analysis of the plasma proteome.J Proteomics 2010;73:629–36.

[28]Keeping AJ,Collins RA.Data variance and statistical significance in 2D-gel electrophoresis and DIGE experiments:comparison of the effects of normalization methods.J Proteome Res 2011;10:1353–60.

[29]Rashwan S,Faheem T,Sarhan A,Youssef BAB.A fuzzywatershed based algorithm for protein spot detection in 2-DGE images.IJCSNS Int J Comput Sci Netw Secur 2010;10:254–60.

[30]Srinark T,Kambhamettu C.An image analysis suite for spot detection and spot matching in two-dimensional electrophoresis gels.Electrophoresis 2008;29:706–15.

[31]Faergestad EM,Rye M,Walczak B,Gidskehaug L,Wold JP,Grove H,et al.Pixel-based analysis of multiple images for the identification of changes:a novel approach applied to unravel proteome patterns of 2-D electrophoresis gel images.Proteomics 2007;7:3450–61.

[32]Eilers PHC. Parametric time warping. Anal Chem 2004;76:404–11.

[33]Kaczmarek K,Walczak B,de Jong S,Vandeginste GM.Baseline reduction in two dimensional gel electrophoresis images.Acta Chromatogr 2005;15:82–96.

[34]Peer P,Corzo LG.Local pixel value collection algorithm for spot segmentation in two-dimensional gel electrophoresis research.Comp Funct Genomics 2007;2007:89596.

[35]Brauner JM,Groemer TW,Stroebel A,Grosse-Holz S,Oberstein T,Wiltfang J,et al.Spot quantification in two dimensional gel electrophoresis image analysis: comparison of different approaches and presentation of a novel compound fitting algorithm.BMC Bioinformatics 2014;15:181.

[36]Piccardi M.Background subtraction techniques:a review.IEEE Inter Conf Syst Man Cybern 2005:3099–104.

[37]Li Y,Li X.A background correction method based on lazy snapping.Proc 7th Int Conf Image Graphics 2013:144–8.

[38]Dowsey AW,English JA,Lisacek F,Morris JS,Yang GZ,Dunn MJ.Image analysis tools and emerging algorithms for expression proteomics.Proteomics 2010;10:4226–57.

[39]Cannistraci CV,Abbas A,Gao X.Median Modified Wiener Filter for nonlinear adaptive spatial denoising of protein NMR multidimensional spectra.Sci Rep 2015;5:8017.

[40]Worz S,Winz ML,Rohr K.Geometric alignment of 2D gel electrophoresis images using physics-based elastic registration.Proc 5th IEEE Int Symp Biomed Imaging:Nano Macro 2008:1135–8.

[41]Sengar RS,Kumar A.Segmentation of two dimensional electrrophoresis gel image using the Wavelet Transform and the Watershed Transform.Proc 2012 National Conf Commun 2012:1–5.

[42]Van Huffel S,Vandewalle J.The total least squares problem:computational aspects and analysis.1st ed.Philadelphia:SIAM;1991.

[43]Daszykowski M,Stanimirova I,Bodzon-Kulakowska A,Silberring J,Lubec G,Walczak B.Start-to-end processing of twodimensional gel electrophoretic images. J Chromatogr 2007;1158:306–17.

[44]Locantore N,Marron JS,Simpson DG,Tripoli N,Zhang JT,Cohen KL,et al.Robust principal component analysis for functional data.Test 1999;8:1–73.

[45]Appel RD,Vargas JR,Palagi PM,Walther D,Hochstrasser DF.Melanie II-a third-generation software package for analysis of two-dimensional electrophoresis images:II.Algorithms.Electrophoresis 1997;18:2735–48.

[46]Do MN,Vetterli M.The contourlet transform:An efficient directional multiresolution image representation.IEEE Trans Image Process 2005;14:2091–106.

[47]Oliveri G,Anselmi N,Massa A.Compressive sensing imaging of non-sparse 2D scatterers by a total-variation approach within the born approximation. IEEE Trans Antennas Propag 2014;62:5157–70.

[48]Tsakanikas,Manolakos.Effective denoising of 2D gel proteomics images using contourlets.Proc 2007 IEEE Int Conf Image 2007:269–72.

[49]Chan TF,Osher S,Shen J.The digital TV filter and nonlinear denoising.IEEE Trans Image Process 2001;10:231–41.

[50]Xin H,Zhao F.Effective denoising methods for two-dimensional gel electrophoresis images.Proc 4th Int Conf Biomed Eng Informatics 2011:1571–4.

[51]Catte´F,Lions PL,Morel JM,Coll T.Image selective smoothing and edge detection by nonlinear diffusion.SIAM J Numer Anal 1992;29:182–93.

[52]Faheem MT,Rashwan S,Sarhan A,Youssef BA.De-noising 2D gel images using wavelet transform based on genetic algorithm.Proc 10th WSEAS Int Conf Appl Comput Appl Comput Sci 2011:187–94.

[53]Bartusek K,Prinosil J,Sme´kal Z.Wavelet-based de-noising techniques in MRI.Comput Methods Programs Biomed 2011;104:480–8.

[54]Munsell BC,Barsanti RJ.Electrophoresis 2D gel residual encoding using adaptive wavelets for image segmentation.Proc IEEE Southeast Conf 2005:163–8.

[55]Pasti L,Walczak B,Massart DL,Reschiglian P.Optimization of signal denoising in discrete wavelet transform.Chemom Intell Lab Syst 1999;48:21–34.

[56]Cheung KW,Po LM.Low complexity preprocessing for 2D discrete multiwavelet transform.Proc 1997 IEEE Int Symp Circuits Syst 1997:0–3.

[57]Mallat S.A Theory for multiresolution signal decomposition:the wavelet representation.Pattern Anal Mach Intell IEEE Trans 1989;11:674–93.

[58]Vora P,Chudasama N.Different image fusion techniques and parameters:a review.Int J Comput Sci Inf Technol 2015;6:889–92.

[59]Chang SG,Yu B,Vetterli M.Adaptive wavelet thresholding for image denoising and compression.IEEE Trans Image Process Publ IEEE Signal Process Soc 2000;9:1532–46.

[60]Barlaud M,Sole P,Gaidon T,Antonini M,Mathieu P.Pyramidal lattice vector quantization for multiscale image coding.IEEE Trans Image Process 1994;3:367–81.

[61]Klein R,Faust O.Genetic algorithms.In:Supply chain management and advanced planning.Berlin:Springer Berlin Heidelberg;2015,p.537–44.

[62]Marengo E,Demartini M,Robotti E,Bobba M.A new algorithm for the simulation of sodium dodecil sulfate two-dimensional polyacrylamide gel electrophoresis data sets.J Proteome Res 2010;9:1864–72.

[63]Gonzalez RC,Woods RE.Digital image processing.2nd ed.New Jersey:Prentice Hall;2002.

[64]Yitzhaky Y,Kopeika NS.Identification of blur parameters from motion blurred images. Graph Model Image Process 1997;59:310–20.

[65]Lester EP,Lemkin PF,Lipkin LE.Protein indexing in leukemias and lymphomas.Ann N Y Acad Sci 1984;428:158–72.

[66]Jonak K,Jedrasiak K,Polan´ski A,Puszyn´ski K.Application of image processing algorithms in proteomics:automatic analysis of 2-d gel electrophoresis images from western blot assay.Proc 2012 Int Conf Comput Vision Graphics 2012:433–40.

[67]Khan A,Yin H.Quality measures for blind image deblurring.Proc 2012 IEEE Int Conf Imaging Syst Tech 2012:456–9.