应用于疾病诊断的图像分析方法
|
摘要
本论文重点讨论用图像分析法来进行疾病诊断。我们运用于演算法中的图像处理技术则集中在三种图像的分类和细分；这三种图像分别为纹理图像,遥感图像,医学图像。这些图像处理技术分别用作小波变换,图像增强,模糊逻辑或者神经网络,特征抽取。对线路检测进行霍夫转换,运用类似于欧几里德距离的数学方法
本论文的目的是为了疾病诊断创造新的算法。但首先我们需要知道在算法中,进行分类和细分的最好的图像处理方法,这使得我们运用多种图像处理方法,其中包括医学图像,以此来确定我们所运用的方法是进行诊断最恰当的方法。
第一个算法有两部分：第一部分是讲对纹理的认识,而第二部分是讲遥感图像。所以在第一部分我们尝试着混用微波和神经网络来寻找适当的方法认识纹理,通过微波,这些纹理会分解成下一等级的图像,这些图像可以分析和提取纹理,通过这些特征我们的神经网络可以用来认识纹理的类型。在这篇论文中运用了五种不同类型的纹理,每种都附有五张图片。在第二部分我们通过参考两个遥感图像分析其纹理。通过运用微波转换和神经网络,我们分别做了分析和分类。我们运用母函数分析了有水,森林和土地的两张图像。该图像是灰度的,尺寸是128×128。该图像的处理出来是把每张图像分成尺寸为32×32的16块。每块都在微波母函数中进行计算,经过分析几个母函数,我们发现(Coifl,Sym5)是最适合的。这些结果是用来进行特征提取(平均,标准偏差,差异),然后将输出作为神经网络的输入。最后从NN中得到结果会得出一个新的纹理类型(森林,土地和水)。
为了分辨不同类型的纹理,我们需要通过如微波转换的分析方法来分析纹理类型。分析的目的是为了找到能够识别纹理的一些特征。这些转换对于提取特征很重要。一般来说,转换过程通过所给信号的旋转和它们的基本功能来体现。这些。。函数一般都是正相交的,这使得转换得以恢复,这意味着原来范围内的原函数在没有信息丢失的情况下就能得到提取。我们运用这种方法得出的结果更精确。
傅立叶变换是回旋变换的一种,该变换通常把信号从时间领域转换到频率领域。由于快速傅立叶变换演算法应用在很广的领域里。这种转换运用快速算法将二维领域拓展。所以,此转换用来进行图像处理。傅立叶变换的下一个形式发展成一个有效转换叫小波变换。
正如快速傅立叶转换一样,离散小波变换很快速,是在数据矢量上的直线操作,该矢量的长度是二者的综合,可以转换成长度相等但数量上不同的矢量。如快速傅立叶转换一样,小波变换是不可避免的,事实上是正交转换—逆转换。小波变换在多分辨率技术中,特别是在最后十年的图像处理中,发挥着很重要的作用。
在离散小波转换的第一步中,信号同步地以样品频率四分之一的截止频率来穿过低压和高压,以此来进行过滤。从高低过滤器中输出是指第一层的大概值和精确值分别的系数。根据乃奎斯特法则,输出信号的频带宽度可以是原频带宽度的一半,并可由两个下降抽样来实现。这个同样的程序可以在第一层的大概系数和精确系数上反复使用,并可得到第二层次的系数。这种分解过程的每一步中,通过过滤频率分解可以增长一倍,通过下降抽样,时间分解变为原来的一半。
通过使用小波转换,我们得到了一些图像特征,通过运用人工神经网络,我们认识了纹理。人工神经网络通过统计法,把真正的工作系统简化了,它是一个信息处理的典范,是受生物神经系统所启发的,诸如大脑,程序信息。人工神经网络的主要目的是是模拟人脑。
网络的高度连接神经计算要素有对输入刺激物有回应,还会学着适应环境。通过网络把一系列输入—输出样品通过,校正重量,来把样品重量调节成正确的价值训练,以此来使网络给予的答案和期望结果之间的误差降至最低。在我们所做的工作中,我们运用了拉凡格—马夸特氏演算法,从而得出了精确地结果。
拉凡格马夸特氏演算法是把演算法用来训练多层次前向网络,是通过使建立在非直线优化技术之上,使平方和的误差减少到最小。拉凡格—马夸特氏演算法是对拉凡格的实践。“是在最小平方方法上对某个难题的解决方法”马夸特“非线性参数的最小平方估计”拉凡格—马夸特演算法被认为是介于最速下降法和高斯-牛顿演算法之间的折中方法。
比起其他两种方法,拉凡格—马夸特演算法拥有更好的收敛性,因其在神经网络的离线训练中是最好的选择而闻名。原因是因为,神经网络最小化问题经常处于不理想状态,拉凡格—马夸特演算法在参数空间里最小限度地影响标准,并忽略琐碎的方向问题。
该论文中用到了五个系列的纹理。每个系列附有五幅图片,都是作为理想状态下的纹理,并有图片尺寸,像素和灰度。第一组用来描述土地纹理,称作(块一块二,块三,块四,块五),第二组用来描述森林的纹理(树一,树二树三,树四,树五),第三组是用来描述草地的,被称作(框架一,框架二,框架三,框架四,框架五),第四组是用来描述羊毛的20幅随意挑选的纹理被用来做试验纹理。
通过使用离散小波转换至第三层,这些纹理将分解成更小的条纹来进行分析。使用到许多小波母函数如(coifl,coif2,Haar,db1,sym2,等等。使用它们后,我们从coifl和sym5小波母函数中提取的要比其他的更好些。同时,我们可以看到,不是所有的纹理中更小的条纹都是有用的,通过使用每组搜集的三种有价值的东西的特征,所以我们只用了有用的小条纹,第一种来自平均值,第二种来自标准差,最后一种来自方差。这些数值收集起来为平均值,标准差和方差所用,而不是作为唯一的数值。这三种数值将放在神经网络中作为输入值,可用来决定纹理的类型。
在训练阶段,人工神经网络已被用来识别依靠输入范围的五纹理类型。神经网络的输入为(3*5=15),因为我们有五组,每组有三个值。输出必须是五组其中之一,但在训练阶段,我们发现出现了一些混合类型,例如(一输出有45%与棉花组织形式相似和55%与羊毛组织相似),因此,在这种情况下,我们通过增加人工神经网络的输出到16个(3*5+1=16),处理像这样的一些错误。通过增加每个输出的百分比解决这个问题,显示出五种纹理类型的相似百分比,例如,从五种类型中之一输出可能会有(3/3(100%)或2/3(70%)或1/3(35%))只有一个混合输出可能包含混合值或与五种类型没有任何的相似性。输出必须在五种纹理类型范围内划定纹理类型和相似百分比,例如,如果输入向量是=[111],输出必须是3/3(100%)来自木材纹理,如果输入向量是=[112],输出必须是2/3(70%)来自木材纹理,如果输入向量是= [123],输出必须是混合纹理。另一种训练神经网络的方法是通过使用从特征提取具有(125)种可能性的输入和输出是(125)种可能性的结合物,因为我们有五种类型,每种类型的纹理具有(25)种可能性。
培训过程中完成了69/250,心方误差(SSE)的20幅图像已被用来测试我们的神经网络,其测试结果是伟大的,并且和测试图像分类结果有95%接近。当我们使用Symb5作为母小波函数,如表4.2所示。从Coifl得到的结果用来识别纹理不合理。Coifl用来Coifl根据从卫星上得到的图片感测图片路线是合理的。第二种算法是宪制性黄疸的诊断,在这个算法中,我们作了诊断黄疸的算法(杜宾-约翰逊,吉尔伯特和转子综合征),该算法是由两部分组成：1)利用小波变换分析图像；通过小波变换,我们收集各种疾病的三个特点。2)通过其直方图为每个图片计算灰阶(白色和黑色的百分比),收集各种疾病的两个特征。总共将有五个值,这五个值将是模糊逻辑预算的输入,用来决定基于这些值的疾病种类。我们给55个遭受不同种类宪制性黄疸的孩童患者作了实验。比起仅仅依靠医生的眼睛诊断,我们的算法得出了更准确的结果。
宪制性是高度特异功能的一种离散步骤,是血胆红素从血液到胆(红黄色颜料,尿液,血液和胆汁)肝途中导致的结果。根据现有的知识,三个主要类型的宪法性黄疸可以区分。
杜宾-约翰逊(道琼斯)综合征是一种常染色体隐性良性状态,其胆红素胆汁分泌有缺陷。缺陷是由于缺少了小管蛋白MRP2位于染色体10q24上,这种物质负责运输葡萄糖醛酸和相关的有机阴离子到胆汁。
从DJ综合征患者的电子显微镜下观察发现,在肝细胞浆有丰富的色素沉淀。该色素颗粒,主要由致密的颗粒物质,但自然的脂质较轻的元素,只有在这些地方可以检测到。
吉尔伯特综合征。这是迄今为止在所有宪制性黄疸症(继承胆红素)中最频繁的一种。这种症出现在青春期,而且其患病率在青少年比例为5—8%。总胆红素一般不超过80μmol/L和直接胆红素低于20%。高胆红素血症环境导致色素积累。色素颗粒仅包含少量的蛋白酶颗粒的丝状部分,其围绕着由脂质球形成的大量的色素体。转子综合症。它和杜宾-Johnson综合征在很多方面都很相似。在在转子综合征发现的色素颗粒的超结构特点包括大约相等的蛋白酶颗粒部分丝状体和脂质球部分。
所有的法师被要求使用具有1000倍放大倍率的电子显微镜,但是从电子显微镜的影像是反面的,所以我们需要获得正常的图像,拍摄的图像以JPEG格式保存有最高分辨率像素的大小,后来称大小为像素。
我们需要预先处理,前处理的目的是提高图像和消除不良影响；图像增强是指其中的任何一个图像的视觉质量得到提高的过程。因为没有图像增强的设置程序所以在图像的获取过程中可能由于不同的光照条件、相机类型等等会有显著的变化。一般来说,增强技术本质上是临时的。
图像调整是第一步,其用来消除噪音(从负转为正常的)带来的图像不想要的部分。我们的工作中需要的是高对比度的图像。在某些情况下,即使图像的动态范围是在显示设备范围内,图像仍然可能是低对比度。这可能是由于图像获取时不良的照明条件或者是捕捉设备较小的动态范围决定的。在这种情况下,简单的像素强度重新调整可能足以提高可视性。这涉及到一个对输入像素强度的分段线性变换
此处的目的是加强低值并且减少像素强度高值,保持完好的像素强度中间值
第二步是对图像进行锐化滤镜处理,通过从自身减去一个模糊的图像版本进行锐化。下一步,我们需要使用分割削减受感染的部分(使用成长切割法).分割过程中,通过从背景组织划定的利益结构和歧视部分,在自动化分析时,做一个重要的承诺。一些有用的算法在分割给了良好的效果,从我们的实验,我们发现其成长切割法对我们的工作非常有益,它是基于以下想法：“细菌从种子像素开始扩散(成长),并尝试占据整个图像。这就是为什么他们称这种方法为'切成长'。细菌滋长和竞争的规则是明显的,在每一离散时间步骤；每一单元试图’进攻部队由袭击者单元的强度定义,在袭击者的和保卫者的特征向量之间的距离”。在每一离散时间里,细菌繁殖和竞争的规则是非常明显的；每一个体试图‘进攻’它的邻居。攻击强度是由袭击者的强决定的,为袭击者的和防卫者的特征向量之间的距离”。
我们也需要用到到图像分析,为了提取最重要的特征,我们需要使用尽可能好的分析方法,因此我们与(DWT)一起尝试,它能获得更准确的结果。
从子波变换中能够得到许多值,我们需要用从中找到有用的特征。特征抽取的目的是从凝聚了类分割信息的图像矩阵中的收集新变量。在获得如同颜色直方图的全球的图像特征或如同形状和纹理的描述符方面,大多数系统把特征抽取作为一个预处理的过程。使用广义的高斯分布,纹理特征被仿造子波系数的边际分布。
在对仅仅含有受感染的部分的图像的进行子波分解后,我们可以提取一些重要的特征,这些特征来自Bior6.8的使用和通过使用平均和标准偏差的Sym4。
然后将那些值或特征输入到模糊集合,以决定这种疾病。使用模糊集合的分类给我们的工作带来了很多好处,因为它提供了很好结果。对于多端输入单输出(MISO)模糊模型系统,有许多类型的模糊神经网络,用于转送多层模糊神经网络,典型的有Mamdani模型、Sugeno模型和Tsukamoto模型,最简单且广泛使用是Sugeno模型,同时也被应用在本文中。为简便起见,我们假定在Sugeno模型的模糊推理系统中有如下三个规则：
规则1：如果STD1是低的,STD2是低的,STD3是低的,PB是低,PW是低的,则f1是D-J。
规则2：如果STD1是中等,STD2是中等,STD3是中等,PB是中等,PW是中等,则f2是G。
规则3：如果STD1是高的,STD2是高的,STD3是高的,PB是高,PW是高的,则f3是R。
当STD1、STD2、STD3、PB(黑色的百分率)和PW(白色的百分率)作为输入参数；低,中间的和高是STD1、STD2、STD3、PB和PW的模糊子集；英尺是模糊规则的结论和输入参数STD1、STD2、STD3、PB和PW的功能；DJ、G和R分别是杜-约二氏、吉尔伯特和转子,如图5-6和表5.1所示的。
我们收集55个案例,这些案例大部分是关于儿童患有不同种类的体质性黄疸这些案例是从位于中国湖南省长沙市的湘雅医院收集的,从1990年到2008年。已进行的实验对医疗有很大的好处并且能够更容易的解决许多问题。
薄基底膜疾病(也被称为良性家族血尿病以及薄基底膜肾病)是大多数无症状血尿的最普通的原因。这种疾病唯一异常的地方就是在肾小球上的基底膜逐渐的消瘦；其定义的超显微结构特征是在没有其他小球变化的情况下肾小球基底膜的消瘦逐渐扩散。通过对薄板内外相对的观察发现肾小球基底膜衰减主要影响致密层。世界卫生组织用来定义薄基膜肾病的肾小球基底膜厚度为成年人小于250nm,2到11岁的小孩为180nm。
我们的工作尝试再现一个或多个医生的意见的性能,最通常在一个特异性诊断中,试图找到一个准确的算法,根据医生鉴定,通过检测膜和计算它的厚度来诊断TBMD。医生将图像放大到5倍以上,更容易计算膜厚度的距离并使用一个标尺测量20个随机点,然后,把结果分开,并放大到比例尺上。然后通过比较来验证其是否正常,对成年人来说与250nm相比,儿童则是180nm。在提出的算法中,超过100个随机点将出现在膜的表面来证明是否正常。在医疗诊断中,系统必须符合几个特殊要求。在这篇文章中我们需要从图像检测膜并删除其它映像内容。许多方法被尝试过,但是要删除其他映像内容而保留膜仍旧非常困难。因此很难找到一个自动化的方法。我们的算法是基于许多图像处理方法,最主要的方法是采用基于内容的图像检索(CBIR),通过在训练阶段使用一个包含许多膜样品的数据库,来只检测膜形状而忽略其他的内容。此外,在训练阶段用一种改进的方法来保持图像,使其具有几乎相同特征。本文提出的能够自动的完成膜的检测并计算膜距离,而不需要人工参与。
我们的算法的目标是开发一个自动计算并检测膜的方法,而这些任务以前都是人工完成的。专家算法或系统的优点是：它提供一致的答案给重复的决定、过程和任务,保持并维持信息的显著性水平,鼓励组织澄清它们的决策的逻辑并从未”忘记”询问,就像人类一样。
在此算法所有的图像都是使用一台6,000X电子显微镜放大得到的；所有的图像都是JPEG格式的,具有最大分辨率544×655个像素,其稍后被转换成到512×512个像素。
为了建立一个数据库需要收集很多图片,数据库中的图片会根据医生的建议给膜着红色,因为我们仅仅需要检测膜并忽视其它图像内容,这些被着色的图片将会用来与原图比较。这些被着色的图片将根据他们原始的结果被分成许多子图(70×70个像素)。我们然后将与原始图像相对应的的所有的包含红颜色(膜)的子图保存到数据库。其他不包含红颜色的部分将被忽视,因为他们不是膜的一部分。膜的不同的形状将被保存在数据库,用来检测膜然后计算膜距离。所有的图像从中国长沙的湘雅医院收集的
根据前面所述,我们的算法主要依靠基于内容的图像检索(CBIR),这需要去收集许多种类的形状作为数据库。
基于内容的图像检索技术是一种自二十世纪九十年代以来的一直在研究先进研究领域。这项技术的目的是通过对用户偏爱性自适应授权以及对图像间相关性和相似性的考虑,从大的多媒体数据库和数字图书馆中搜索图像。这可以通过测试的用户的对检索结果反应中获得,检索结果都是以反馈的形式出现的。CBR依靠对原始的特征的描述,例如颜色,形状,和纹理,这些可以图像本身自动地被提取出来。这些原始的特征被列入IR,用来补救使用基于文本图像检索技术出现的问题。
基于内容的图像检索(CBIR的),过去的十年中,已经在大范围探索。在C BIR的情况下,图像是由一个低级别的视觉特征,它们没有什么高层次的语义概念直接相关集,在高层次概念和低层次特征之间的差距是制约CBIR系统发展的最主要的困难。相关反馈和区域的图像检索(RBIR)已经被提出来弥补这一差距。相关反馈机制是一个反复的学习过程,它一般被视为网上监督学习。
在创建数据库的不同的膜的形状之后,子图像(膜)被转换为灰度级图片,这些图片是在不同年份收集的,并且通过使用一种图像加强方法,加强相同的性质我们尝试了许多增强方法。最后,通过实验确定,得出的调整后的高对比度的图像是很必要的。用来消除由于噪声(从负转为正常的)带来不想要部分的图像的图像调整是第一步。高对比度的图像,也是我们的工作需要的。我们需要使用55平均噪声过滤来删除来自子图像的噪声和小结构,仅仅保护膜。
Hough变换是一个强大的全球性的边缘检测方法。它在笛卡尔空间坐标和参数空间中转换,在参数空间,直线(或其他边界方程)能够被定义,此种变换用在对形状检测的各种相关方法中。
为了检测中的子图像的边界膜,有必要首先检测了膜的边缘。通过使用Can ny过滤器,子图像将被转换成二进制图像。在尝试过许多边缘检测滤波器之后,得出Canny算法滤波器得出最好的结果。为了找出膜的表面,并且轻松地计算出膜的距离,有必要知道在两对膜表面的对应边的距离。因此,直线检测方法是必须的,超过四条直线可得出结论,因此很容易计算出膜的距离。
通过欧几里得距离计算在相关的两条线(这两条线在膜表面的相关的两边)之间的两点的距离,计算膜的厚度,并且确定两个点,我们需要知道两行的每个斜坡。
最终值以像素为单位计算,因此,测量单位必须从像素转换成纳米,因为纳米是正常与不正常相比较的标准单位。因为所有的相片都是用6000倍放大,横向和纵向的长度都是300dpi,厘米=纳米的情况下捕捉的。
实验为10个新的图片进行了计算,并且所得的结果非常接近手动计算的结果,只有很小的百分比误差。十张图片都是从湘雅第一医院(长沙,中国)获得的,其中四张是异常的孩童图片,四张异常的成人图片,还有两张是正常的成年人。运用之前提出的方法,我们尝试了许多不同的形状膜,但这个算法还是面临着一些问题。在膜的相同边,只有两条线,而且我们需要计算两对应边两条线之间的距离。另一个问题是,这两条线并不完全在同一侧,但是当他们的正交线相交,虽然该算法计算了其距离,但结果是不准确的,它可能因此降低了本来值就很小的膜厚度。
因此,我们提出了两个条件：第一是不能计算在膜相同边的或者在忽略了最小距离后不平行的两条线的距离；第二个条件是保证两者之间的对应线的夹角至小
因此,小的距离可以忽略,该算法允许去计算最小距离,这可能会出现错误的结果。归根到底,我们的算法具有较人工方法的准确性,因为它随机选择点(如150)要比手工方法多,手工方法只选择20个随机点计算的平均值。
我们的结论是,在特征提取方面,如根据平均数,STD与方差作为关键来分类,实验结果表明平均数,STD与方差的分类比最大、最小和中值要好。我们也试图使用最大、最小值和中值作为特征提取来分类,但结果并不是很好,比如说最大值可以在170—360之间,同样也可在160-350之间,这没什么用,同样,在最少值和中值也存在同样的问题
在图像分析中分别使用小波分析(哈尔,Symmelet2,Symmelet5, Coifletl ,Coiflet2和Daubechies),因为Daubechies和Haar具有相同的系数,Daubechies分析的结果和Haar分析的结果有相同的。当我们使用的图片是从普通摄像机拍摄而不是从卫星拍摄(遥感图片)时,Sym5具有更好的结果,其他的则是coifl最好的。我们使用的第三层次的WT将每张图片分解成12个子图片；实验观察到子频带没什么用,这意味着提取的特征不适合分类,就像下面一样：
在我们分类的进程中,平均上所有子频带(从LL1至HH3)都是有用的如果STDLL1, LL2,和LL3没什么用处,可能会导致在分类过程中出现问题,有些是都是有用的,方差HH1, HH2, HH3, HL1和LH1这些在分类过程中非常有用。(提示：LL1的意思是在第一级的近似系数和HH指细节系数等)
在第四章的第二部分,我们的实验表明,coiflis用在遥感领域非常有效地,因为它执行从其他过滤器获得的结果。但是,如果我们要处理的天然纹理(即不属于远程图像),我们可以使用Sym5。该Sym5比(哈尔,Symmelet2, Coiflet2和Da ube-chiesl)更好。使用这种方法作为一个(FE),我们发现,只有LL1, LL2, LL3是正的并且有比较大的值,但其他子带都是负值。然而,使用(性病)作为(FE)作出只LL1,LL2和LL3不被用于作为分类特征,这是因为在这些子带中,其值比其他正值更大,而且这可能会造成各种纹理类型互相干扰。在这里,HL1,HH2和HH3是用来功能分类的,因为只有这些子带是稳定的。
在小波家族,我们尝试了很多方法加强和特征提取。而且我们发现：小波家族;Bior6.8和Symb4是我们实验最好的分析手段,为了三个层次通过使用近似部分的WT,我们得到了最好的结果。
这种方法的第二个算法中,图像调整是我们算法最好的增强方法,其主要是为了取代手工方法,因为这种算法被证明是更有效和更精确。它还有另一个优势,就是,任何医生可以轻松地使用它。更进一步说,该方法不需要医生诊断专家的诊断是否正常或不正常的,它可以在任何时间和任何地点,不像手动方法需要特殊设备。
我们的未来工作分为三部：；第一部分是自然图像或遥感图像的分类。我们采用子波变换来分析,用神经网络来分类,也许其他研究人员会采用其他的方法来分析并分类,比如采用GA或模糊逻辑FZ。
在第二个算法中我们采用模糊逻辑,其他研究人员可能尝试其他方法,例如使用彩色图像使得这三种疾病之间的不同点变得更清晰。
在第三个算法中不同的研究人员可能朝不同的方向尝试着去诊断,我们采用的是数学方法,其他人科恩那个采用蚁群算法。
Our dissertation focus on image analysis methods for disease diagnosis, the image processing techniques which we need to use in implementation of our algorithms is focus on classification and segmentation for three types of images; texture images, remote sens-ing images and medical images. The image processing techniques which used as:wavelet transform, image enhancement, fuzzy logic or neural network, feature extraction, Hough transform for line detection and mathematic methods like Euclidean distance.
The aim of our dissertation is to make new algorithms for diseases diagnosis, but first we need to know the best image methods for classification and segmentation in our algorithms which made us to try to use many kinds of images including medical images, to make sure that our way is best way for diagnosis.
We made three algorithms for image analysis to reach the disease diagnosis; one is for general image classification that means we use normal and remote sensing images and the two others is for disease diagnosis that means using medical images. The first al-gorithm is having two parts; the first part is for recognition of textures and the second one is for remote sensing images. So in the first part we try to mix between wavelet and neu-ral network to find suitable way for recognition the textures, via wavelet the texture will decompose into sub-images these sub-images will analyse and extract features, these fea-tures will be the key of the neural to recognition the different types of texture, via these features our neural network will recognition the type of the texture. In our work five dif-ferent types of texture have been used, each type has five pictures. The results from our way had more accuracy. In the second part we analysis textures of remote sensing images by taking two reference remote sensing images. By employ the wavelet transform and neural network for analysis and classification respectively. We use (symmlet5) and (ciofletl) mother functions for analyzing the two images that contains water, forest and earth. The images are gray level and (128×128) size. The processing is carried out to di-vide each image into (16) blocks with size (32×32). Each block will be entered to the wavelet mother function, after trying several mother functions, we found that the (Coifl, Sym5) are the best choice. The results are passed to the features extraction (mean, stan-dard deviation, and variance) and the output is then fed as input to the neural network (NN). Finally the result from NN with (Levenberg Marquardt (LM) algorithm) gives the type of texture (forest, earth, and water).
The Second algorithm is for Constitutional Jaundice diagnosis, in this algorithm we have made an algorithm to diagnose the constitutional jaundice (Dubin-Johnson, Gil-bert and Rotor syndrome) the algorithm is decomposed into two parts:1) Using wavelet transform to analyse the image; via wavelet transform we collected three features for each kind of disease 2) Calculates the percentage of the gray scales (percentage of white and black colour) for each image via its histogram, it collects two features for each kind of disease. In total there will be five values; these five values will be the inputs for the fuzzy logic that will decide the kind of disease based on these values. We made experiments for 55 cases mostly for children who suffered from different kinds of constitutional jaundice. Our algorithm yields more accurate results compared to the diagnosis by a doctor's eyes only. We collected 55 cases mostly for children that suffered from different kinds of the Constitutional Jaundice.
The third algorithm is to diagnose thin basement membrane. The idea of our algo-rithm is based on content based image retrieval. The diagnosis of this disease is depend on measure of the thickness of the membrane, the traditional way for diagnosis is to manually calculate the thickness, so we suggest an automatic algorithm to detect the membrane and calculate thickness, at first the detecting of membrane is decided by doc-tor diagnosis, then we used some pictures to built database, this database will detect the membrane automatically then calculate the thickness to know whether it is normal or ab-normal. Compared with manually ways our algorithm is easy to use and has more accu-racy.
本论文的目的是为了疾病诊断创造新的算法。但首先我们需要知道在算法中,进行分类和细分的最好的图像处理方法,这使得我们运用多种图像处理方法,其中包括医学图像,以此来确定我们所运用的方法是进行诊断最恰当的方法。
第一个算法有两部分：第一部分是讲对纹理的认识,而第二部分是讲遥感图像。所以在第一部分我们尝试着混用微波和神经网络来寻找适当的方法认识纹理,通过微波,这些纹理会分解成下一等级的图像,这些图像可以分析和提取纹理,通过这些特征我们的神经网络可以用来认识纹理的类型。在这篇论文中运用了五种不同类型的纹理,每种都附有五张图片。在第二部分我们通过参考两个遥感图像分析其纹理。通过运用微波转换和神经网络,我们分别做了分析和分类。我们运用母函数分析了有水,森林和土地的两张图像。该图像是灰度的,尺寸是128×128。该图像的处理出来是把每张图像分成尺寸为32×32的16块。每块都在微波母函数中进行计算,经过分析几个母函数,我们发现(Coifl,Sym5)是最适合的。这些结果是用来进行特征提取(平均,标准偏差,差异),然后将输出作为神经网络的输入。最后从NN中得到结果会得出一个新的纹理类型(森林,土地和水)。
为了分辨不同类型的纹理,我们需要通过如微波转换的分析方法来分析纹理类型。分析的目的是为了找到能够识别纹理的一些特征。这些转换对于提取特征很重要。一般来说,转换过程通过所给信号的旋转和它们的基本功能来体现。这些。。函数一般都是正相交的,这使得转换得以恢复,这意味着原来范围内的原函数在没有信息丢失的情况下就能得到提取。我们运用这种方法得出的结果更精确。
傅立叶变换是回旋变换的一种,该变换通常把信号从时间领域转换到频率领域。由于快速傅立叶变换演算法应用在很广的领域里。这种转换运用快速算法将二维领域拓展。所以,此转换用来进行图像处理。傅立叶变换的下一个形式发展成一个有效转换叫小波变换。
正如快速傅立叶转换一样,离散小波变换很快速,是在数据矢量上的直线操作,该矢量的长度是二者的综合,可以转换成长度相等但数量上不同的矢量。如快速傅立叶转换一样,小波变换是不可避免的,事实上是正交转换—逆转换。小波变换在多分辨率技术中,特别是在最后十年的图像处理中,发挥着很重要的作用。
在离散小波转换的第一步中,信号同步地以样品频率四分之一的截止频率来穿过低压和高压,以此来进行过滤。从高低过滤器中输出是指第一层的大概值和精确值分别的系数。根据乃奎斯特法则,输出信号的频带宽度可以是原频带宽度的一半,并可由两个下降抽样来实现。这个同样的程序可以在第一层的大概系数和精确系数上反复使用,并可得到第二层次的系数。这种分解过程的每一步中,通过过滤频率分解可以增长一倍,通过下降抽样,时间分解变为原来的一半。
通过使用小波转换,我们得到了一些图像特征,通过运用人工神经网络,我们认识了纹理。人工神经网络通过统计法,把真正的工作系统简化了,它是一个信息处理的典范,是受生物神经系统所启发的,诸如大脑,程序信息。人工神经网络的主要目的是是模拟人脑。
网络的高度连接神经计算要素有对输入刺激物有回应,还会学着适应环境。通过网络把一系列输入—输出样品通过,校正重量,来把样品重量调节成正确的价值训练,以此来使网络给予的答案和期望结果之间的误差降至最低。在我们所做的工作中,我们运用了拉凡格—马夸特氏演算法,从而得出了精确地结果。
拉凡格马夸特氏演算法是把演算法用来训练多层次前向网络,是通过使建立在非直线优化技术之上,使平方和的误差减少到最小。拉凡格—马夸特氏演算法是对拉凡格的实践。“是在最小平方方法上对某个难题的解决方法”马夸特“非线性参数的最小平方估计”拉凡格—马夸特演算法被认为是介于最速下降法和高斯-牛顿演算法之间的折中方法。
比起其他两种方法,拉凡格—马夸特演算法拥有更好的收敛性,因其在神经网络的离线训练中是最好的选择而闻名。原因是因为,神经网络最小化问题经常处于不理想状态,拉凡格—马夸特演算法在参数空间里最小限度地影响标准,并忽略琐碎的方向问题。
该论文中用到了五个系列的纹理。每个系列附有五幅图片,都是作为理想状态下的纹理,并有图片尺寸,像素和灰度。第一组用来描述土地纹理,称作(块一块二,块三,块四,块五),第二组用来描述森林的纹理(树一,树二树三,树四,树五),第三组是用来描述草地的,被称作(框架一,框架二,框架三,框架四,框架五),第四组是用来描述羊毛的20幅随意挑选的纹理被用来做试验纹理。
通过使用离散小波转换至第三层,这些纹理将分解成更小的条纹来进行分析。使用到许多小波母函数如(coifl,coif2,Haar,db1,sym2,等等。使用它们后,我们从coifl和sym5小波母函数中提取的要比其他的更好些。同时,我们可以看到,不是所有的纹理中更小的条纹都是有用的,通过使用每组搜集的三种有价值的东西的特征,所以我们只用了有用的小条纹,第一种来自平均值,第二种来自标准差,最后一种来自方差。这些数值收集起来为平均值,标准差和方差所用,而不是作为唯一的数值。这三种数值将放在神经网络中作为输入值,可用来决定纹理的类型。
在训练阶段,人工神经网络已被用来识别依靠输入范围的五纹理类型。神经网络的输入为(3*5=15),因为我们有五组,每组有三个值。输出必须是五组其中之一,但在训练阶段,我们发现出现了一些混合类型,例如(一输出有45%与棉花组织形式相似和55%与羊毛组织相似),因此,在这种情况下,我们通过增加人工神经网络的输出到16个(3*5+1=16),处理像这样的一些错误。通过增加每个输出的百分比解决这个问题,显示出五种纹理类型的相似百分比,例如,从五种类型中之一输出可能会有(3/3(100%)或2/3(70%)或1/3(35%))只有一个混合输出可能包含混合值或与五种类型没有任何的相似性。输出必须在五种纹理类型范围内划定纹理类型和相似百分比,例如,如果输入向量是=[111],输出必须是3/3(100%)来自木材纹理,如果输入向量是=[112],输出必须是2/3(70%)来自木材纹理,如果输入向量是= [123],输出必须是混合纹理。另一种训练神经网络的方法是通过使用从特征提取具有(125)种可能性的输入和输出是(125)种可能性的结合物,因为我们有五种类型,每种类型的纹理具有(25)种可能性。
培训过程中完成了69/250,心方误差(SSE)的20幅图像已被用来测试我们的神经网络,其测试结果是伟大的,并且和测试图像分类结果有95%接近。当我们使用Symb5作为母小波函数,如表4.2所示。从Coifl得到的结果用来识别纹理不合理。Coifl用来Coifl根据从卫星上得到的图片感测图片路线是合理的。第二种算法是宪制性黄疸的诊断,在这个算法中,我们作了诊断黄疸的算法(杜宾-约翰逊,吉尔伯特和转子综合征),该算法是由两部分组成：1)利用小波变换分析图像；通过小波变换,我们收集各种疾病的三个特点。2)通过其直方图为每个图片计算灰阶(白色和黑色的百分比),收集各种疾病的两个特征。总共将有五个值,这五个值将是模糊逻辑预算的输入,用来决定基于这些值的疾病种类。我们给55个遭受不同种类宪制性黄疸的孩童患者作了实验。比起仅仅依靠医生的眼睛诊断,我们的算法得出了更准确的结果。
宪制性是高度特异功能的一种离散步骤,是血胆红素从血液到胆(红黄色颜料,尿液,血液和胆汁)肝途中导致的结果。根据现有的知识,三个主要类型的宪法性黄疸可以区分。
杜宾-约翰逊(道琼斯)综合征是一种常染色体隐性良性状态,其胆红素胆汁分泌有缺陷。缺陷是由于缺少了小管蛋白MRP2位于染色体10q24上,这种物质负责运输葡萄糖醛酸和相关的有机阴离子到胆汁。
从DJ综合征患者的电子显微镜下观察发现,在肝细胞浆有丰富的色素沉淀。该色素颗粒,主要由致密的颗粒物质,但自然的脂质较轻的元素,只有在这些地方可以检测到。
吉尔伯特综合征。这是迄今为止在所有宪制性黄疸症(继承胆红素)中最频繁的一种。这种症出现在青春期,而且其患病率在青少年比例为5—8%。总胆红素一般不超过80μmol/L和直接胆红素低于20%。高胆红素血症环境导致色素积累。色素颗粒仅包含少量的蛋白酶颗粒的丝状部分,其围绕着由脂质球形成的大量的色素体。转子综合症。它和杜宾-Johnson综合征在很多方面都很相似。在在转子综合征发现的色素颗粒的超结构特点包括大约相等的蛋白酶颗粒部分丝状体和脂质球部分。
所有的法师被要求使用具有1000倍放大倍率的电子显微镜,但是从电子显微镜的影像是反面的,所以我们需要获得正常的图像,拍摄的图像以JPEG格式保存有最高分辨率像素的大小,后来称大小为像素。
我们需要预先处理,前处理的目的是提高图像和消除不良影响；图像增强是指其中的任何一个图像的视觉质量得到提高的过程。因为没有图像增强的设置程序所以在图像的获取过程中可能由于不同的光照条件、相机类型等等会有显著的变化。一般来说,增强技术本质上是临时的。
图像调整是第一步,其用来消除噪音(从负转为正常的)带来的图像不想要的部分。我们的工作中需要的是高对比度的图像。在某些情况下,即使图像的动态范围是在显示设备范围内,图像仍然可能是低对比度。这可能是由于图像获取时不良的照明条件或者是捕捉设备较小的动态范围决定的。在这种情况下,简单的像素强度重新调整可能足以提高可视性。这涉及到一个对输入像素强度的分段线性变换
此处的目的是加强低值并且减少像素强度高值,保持完好的像素强度中间值
第二步是对图像进行锐化滤镜处理,通过从自身减去一个模糊的图像版本进行锐化。下一步,我们需要使用分割削减受感染的部分(使用成长切割法).分割过程中,通过从背景组织划定的利益结构和歧视部分,在自动化分析时,做一个重要的承诺。一些有用的算法在分割给了良好的效果,从我们的实验,我们发现其成长切割法对我们的工作非常有益,它是基于以下想法：“细菌从种子像素开始扩散(成长),并尝试占据整个图像。这就是为什么他们称这种方法为'切成长'。细菌滋长和竞争的规则是明显的,在每一离散时间步骤；每一单元试图’进攻部队由袭击者单元的强度定义,在袭击者的和保卫者的特征向量之间的距离”。在每一离散时间里,细菌繁殖和竞争的规则是非常明显的；每一个体试图‘进攻’它的邻居。攻击强度是由袭击者的强决定的,为袭击者的和防卫者的特征向量之间的距离”。
我们也需要用到到图像分析,为了提取最重要的特征,我们需要使用尽可能好的分析方法,因此我们与(DWT)一起尝试,它能获得更准确的结果。
从子波变换中能够得到许多值,我们需要用从中找到有用的特征。特征抽取的目的是从凝聚了类分割信息的图像矩阵中的收集新变量。在获得如同颜色直方图的全球的图像特征或如同形状和纹理的描述符方面,大多数系统把特征抽取作为一个预处理的过程。使用广义的高斯分布,纹理特征被仿造子波系数的边际分布。
在对仅仅含有受感染的部分的图像的进行子波分解后,我们可以提取一些重要的特征,这些特征来自Bior6.8的使用和通过使用平均和标准偏差的Sym4。
然后将那些值或特征输入到模糊集合,以决定这种疾病。使用模糊集合的分类给我们的工作带来了很多好处,因为它提供了很好结果。对于多端输入单输出(MISO)模糊模型系统,有许多类型的模糊神经网络,用于转送多层模糊神经网络,典型的有Mamdani模型、Sugeno模型和Tsukamoto模型,最简单且广泛使用是Sugeno模型,同时也被应用在本文中。为简便起见,我们假定在Sugeno模型的模糊推理系统中有如下三个规则：
规则1：如果STD1是低的,STD2是低的,STD3是低的,PB是低,PW是低的,则f1是D-J。
规则2：如果STD1是中等,STD2是中等,STD3是中等,PB是中等,PW是中等,则f2是G。
规则3：如果STD1是高的,STD2是高的,STD3是高的,PB是高,PW是高的,则f3是R。
当STD1、STD2、STD3、PB(黑色的百分率)和PW(白色的百分率)作为输入参数；低,中间的和高是STD1、STD2、STD3、PB和PW的模糊子集；英尺是模糊规则的结论和输入参数STD1、STD2、STD3、PB和PW的功能；DJ、G和R分别是杜-约二氏、吉尔伯特和转子,如图5-6和表5.1所示的。
我们收集55个案例,这些案例大部分是关于儿童患有不同种类的体质性黄疸这些案例是从位于中国湖南省长沙市的湘雅医院收集的,从1990年到2008年。已进行的实验对医疗有很大的好处并且能够更容易的解决许多问题。
薄基底膜疾病(也被称为良性家族血尿病以及薄基底膜肾病)是大多数无症状血尿的最普通的原因。这种疾病唯一异常的地方就是在肾小球上的基底膜逐渐的消瘦；其定义的超显微结构特征是在没有其他小球变化的情况下肾小球基底膜的消瘦逐渐扩散。通过对薄板内外相对的观察发现肾小球基底膜衰减主要影响致密层。世界卫生组织用来定义薄基膜肾病的肾小球基底膜厚度为成年人小于250nm,2到11岁的小孩为180nm。
我们的工作尝试再现一个或多个医生的意见的性能,最通常在一个特异性诊断中,试图找到一个准确的算法,根据医生鉴定,通过检测膜和计算它的厚度来诊断TBMD。医生将图像放大到5倍以上,更容易计算膜厚度的距离并使用一个标尺测量20个随机点,然后,把结果分开,并放大到比例尺上。然后通过比较来验证其是否正常,对成年人来说与250nm相比,儿童则是180nm。在提出的算法中,超过100个随机点将出现在膜的表面来证明是否正常。在医疗诊断中,系统必须符合几个特殊要求。在这篇文章中我们需要从图像检测膜并删除其它映像内容。许多方法被尝试过,但是要删除其他映像内容而保留膜仍旧非常困难。因此很难找到一个自动化的方法。我们的算法是基于许多图像处理方法,最主要的方法是采用基于内容的图像检索(CBIR),通过在训练阶段使用一个包含许多膜样品的数据库,来只检测膜形状而忽略其他的内容。此外,在训练阶段用一种改进的方法来保持图像,使其具有几乎相同特征。本文提出的能够自动的完成膜的检测并计算膜距离,而不需要人工参与。
我们的算法的目标是开发一个自动计算并检测膜的方法,而这些任务以前都是人工完成的。专家算法或系统的优点是：它提供一致的答案给重复的决定、过程和任务,保持并维持信息的显著性水平,鼓励组织澄清它们的决策的逻辑并从未”忘记”询问,就像人类一样。
在此算法所有的图像都是使用一台6,000X电子显微镜放大得到的；所有的图像都是JPEG格式的,具有最大分辨率544×655个像素,其稍后被转换成到512×512个像素。
为了建立一个数据库需要收集很多图片,数据库中的图片会根据医生的建议给膜着红色,因为我们仅仅需要检测膜并忽视其它图像内容,这些被着色的图片将会用来与原图比较。这些被着色的图片将根据他们原始的结果被分成许多子图(70×70个像素)。我们然后将与原始图像相对应的的所有的包含红颜色(膜)的子图保存到数据库。其他不包含红颜色的部分将被忽视,因为他们不是膜的一部分。膜的不同的形状将被保存在数据库,用来检测膜然后计算膜距离。所有的图像从中国长沙的湘雅医院收集的
根据前面所述,我们的算法主要依靠基于内容的图像检索(CBIR),这需要去收集许多种类的形状作为数据库。
基于内容的图像检索技术是一种自二十世纪九十年代以来的一直在研究先进研究领域。这项技术的目的是通过对用户偏爱性自适应授权以及对图像间相关性和相似性的考虑,从大的多媒体数据库和数字图书馆中搜索图像。这可以通过测试的用户的对检索结果反应中获得,检索结果都是以反馈的形式出现的。CBR依靠对原始的特征的描述,例如颜色,形状,和纹理,这些可以图像本身自动地被提取出来。这些原始的特征被列入IR,用来补救使用基于文本图像检索技术出现的问题。
基于内容的图像检索(CBIR的),过去的十年中,已经在大范围探索。在C BIR的情况下,图像是由一个低级别的视觉特征,它们没有什么高层次的语义概念直接相关集,在高层次概念和低层次特征之间的差距是制约CBIR系统发展的最主要的困难。相关反馈和区域的图像检索(RBIR)已经被提出来弥补这一差距。相关反馈机制是一个反复的学习过程,它一般被视为网上监督学习。
在创建数据库的不同的膜的形状之后,子图像(膜)被转换为灰度级图片,这些图片是在不同年份收集的,并且通过使用一种图像加强方法,加强相同的性质我们尝试了许多增强方法。最后,通过实验确定,得出的调整后的高对比度的图像是很必要的。用来消除由于噪声(从负转为正常的)带来不想要部分的图像的图像调整是第一步。高对比度的图像,也是我们的工作需要的。我们需要使用55平均噪声过滤来删除来自子图像的噪声和小结构,仅仅保护膜。
Hough变换是一个强大的全球性的边缘检测方法。它在笛卡尔空间坐标和参数空间中转换,在参数空间,直线(或其他边界方程)能够被定义,此种变换用在对形状检测的各种相关方法中。
为了检测中的子图像的边界膜,有必要首先检测了膜的边缘。通过使用Can ny过滤器,子图像将被转换成二进制图像。在尝试过许多边缘检测滤波器之后,得出Canny算法滤波器得出最好的结果。为了找出膜的表面,并且轻松地计算出膜的距离,有必要知道在两对膜表面的对应边的距离。因此,直线检测方法是必须的,超过四条直线可得出结论,因此很容易计算出膜的距离。
通过欧几里得距离计算在相关的两条线(这两条线在膜表面的相关的两边)之间的两点的距离,计算膜的厚度,并且确定两个点,我们需要知道两行的每个斜坡。
最终值以像素为单位计算,因此,测量单位必须从像素转换成纳米,因为纳米是正常与不正常相比较的标准单位。因为所有的相片都是用6000倍放大,横向和纵向的长度都是300dpi,厘米=纳米的情况下捕捉的。
实验为10个新的图片进行了计算,并且所得的结果非常接近手动计算的结果,只有很小的百分比误差。十张图片都是从湘雅第一医院(长沙,中国)获得的,其中四张是异常的孩童图片,四张异常的成人图片,还有两张是正常的成年人。运用之前提出的方法,我们尝试了许多不同的形状膜,但这个算法还是面临着一些问题。在膜的相同边,只有两条线,而且我们需要计算两对应边两条线之间的距离。另一个问题是,这两条线并不完全在同一侧,但是当他们的正交线相交,虽然该算法计算了其距离,但结果是不准确的,它可能因此降低了本来值就很小的膜厚度。
因此,我们提出了两个条件：第一是不能计算在膜相同边的或者在忽略了最小距离后不平行的两条线的距离；第二个条件是保证两者之间的对应线的夹角至小
因此,小的距离可以忽略,该算法允许去计算最小距离,这可能会出现错误的结果。归根到底,我们的算法具有较人工方法的准确性,因为它随机选择点(如150)要比手工方法多,手工方法只选择20个随机点计算的平均值。
我们的结论是,在特征提取方面,如根据平均数,STD与方差作为关键来分类,实验结果表明平均数,STD与方差的分类比最大、最小和中值要好。我们也试图使用最大、最小值和中值作为特征提取来分类,但结果并不是很好,比如说最大值可以在170—360之间,同样也可在160-350之间,这没什么用,同样,在最少值和中值也存在同样的问题
在图像分析中分别使用小波分析(哈尔,Symmelet2,Symmelet5, Coifletl ,Coiflet2和Daubechies),因为Daubechies和Haar具有相同的系数,Daubechies分析的结果和Haar分析的结果有相同的。当我们使用的图片是从普通摄像机拍摄而不是从卫星拍摄(遥感图片)时,Sym5具有更好的结果,其他的则是coifl最好的。我们使用的第三层次的WT将每张图片分解成12个子图片；实验观察到子频带没什么用,这意味着提取的特征不适合分类,就像下面一样：
在我们分类的进程中,平均上所有子频带(从LL1至HH3)都是有用的如果STDLL1, LL2,和LL3没什么用处,可能会导致在分类过程中出现问题,有些是都是有用的,方差HH1, HH2, HH3, HL1和LH1这些在分类过程中非常有用。(提示：LL1的意思是在第一级的近似系数和HH指细节系数等)
在第四章的第二部分,我们的实验表明,coiflis用在遥感领域非常有效地,因为它执行从其他过滤器获得的结果。但是,如果我们要处理的天然纹理(即不属于远程图像),我们可以使用Sym5。该Sym5比(哈尔,Symmelet2, Coiflet2和Da ube-chiesl)更好。使用这种方法作为一个(FE),我们发现,只有LL1, LL2, LL3是正的并且有比较大的值,但其他子带都是负值。然而,使用(性病)作为(FE)作出只LL1,LL2和LL3不被用于作为分类特征,这是因为在这些子带中,其值比其他正值更大,而且这可能会造成各种纹理类型互相干扰。在这里,HL1,HH2和HH3是用来功能分类的,因为只有这些子带是稳定的。
在小波家族,我们尝试了很多方法加强和特征提取。而且我们发现：小波家族;Bior6.8和Symb4是我们实验最好的分析手段,为了三个层次通过使用近似部分的WT,我们得到了最好的结果。
这种方法的第二个算法中,图像调整是我们算法最好的增强方法,其主要是为了取代手工方法,因为这种算法被证明是更有效和更精确。它还有另一个优势,就是,任何医生可以轻松地使用它。更进一步说,该方法不需要医生诊断专家的诊断是否正常或不正常的,它可以在任何时间和任何地点,不像手动方法需要特殊设备。
我们的未来工作分为三部：；第一部分是自然图像或遥感图像的分类。我们采用子波变换来分析,用神经网络来分类,也许其他研究人员会采用其他的方法来分析并分类,比如采用GA或模糊逻辑FZ。
在第二个算法中我们采用模糊逻辑,其他研究人员可能尝试其他方法,例如使用彩色图像使得这三种疾病之间的不同点变得更清晰。
在第三个算法中不同的研究人员可能朝不同的方向尝试着去诊断,我们采用的是数学方法,其他人科恩那个采用蚁群算法。
Our dissertation focus on image analysis methods for disease diagnosis, the image processing techniques which we need to use in implementation of our algorithms is focus on classification and segmentation for three types of images; texture images, remote sens-ing images and medical images. The image processing techniques which used as:wavelet transform, image enhancement, fuzzy logic or neural network, feature extraction, Hough transform for line detection and mathematic methods like Euclidean distance.
The aim of our dissertation is to make new algorithms for diseases diagnosis, but first we need to know the best image methods for classification and segmentation in our algorithms which made us to try to use many kinds of images including medical images, to make sure that our way is best way for diagnosis.
We made three algorithms for image analysis to reach the disease diagnosis; one is for general image classification that means we use normal and remote sensing images and the two others is for disease diagnosis that means using medical images. The first al-gorithm is having two parts; the first part is for recognition of textures and the second one is for remote sensing images. So in the first part we try to mix between wavelet and neu-ral network to find suitable way for recognition the textures, via wavelet the texture will decompose into sub-images these sub-images will analyse and extract features, these fea-tures will be the key of the neural to recognition the different types of texture, via these features our neural network will recognition the type of the texture. In our work five dif-ferent types of texture have been used, each type has five pictures. The results from our way had more accuracy. In the second part we analysis textures of remote sensing images by taking two reference remote sensing images. By employ the wavelet transform and neural network for analysis and classification respectively. We use (symmlet5) and (ciofletl) mother functions for analyzing the two images that contains water, forest and earth. The images are gray level and (128×128) size. The processing is carried out to di-vide each image into (16) blocks with size (32×32). Each block will be entered to the wavelet mother function, after trying several mother functions, we found that the (Coifl, Sym5) are the best choice. The results are passed to the features extraction (mean, stan-dard deviation, and variance) and the output is then fed as input to the neural network (NN). Finally the result from NN with (Levenberg Marquardt (LM) algorithm) gives the type of texture (forest, earth, and water).
The Second algorithm is for Constitutional Jaundice diagnosis, in this algorithm we have made an algorithm to diagnose the constitutional jaundice (Dubin-Johnson, Gil-bert and Rotor syndrome) the algorithm is decomposed into two parts:1) Using wavelet transform to analyse the image; via wavelet transform we collected three features for each kind of disease 2) Calculates the percentage of the gray scales (percentage of white and black colour) for each image via its histogram, it collects two features for each kind of disease. In total there will be five values; these five values will be the inputs for the fuzzy logic that will decide the kind of disease based on these values. We made experiments for 55 cases mostly for children who suffered from different kinds of constitutional jaundice. Our algorithm yields more accurate results compared to the diagnosis by a doctor's eyes only. We collected 55 cases mostly for children that suffered from different kinds of the Constitutional Jaundice.
The third algorithm is to diagnose thin basement membrane. The idea of our algo-rithm is based on content based image retrieval. The diagnosis of this disease is depend on measure of the thickness of the membrane, the traditional way for diagnosis is to manually calculate the thickness, so we suggest an automatic algorithm to detect the membrane and calculate thickness, at first the detecting of membrane is decided by doc-tor diagnosis, then we used some pictures to built database, this database will detect the membrane automatically then calculate the thickness to know whether it is normal or ab-normal. Compared with manually ways our algorithm is easy to use and has more accu-racy.
引文
[1]Ali S K, Zou B J, Peng L X. A Wavelet-fuzzy approach for diagnosis the consti-tutional jaundice. In:Proceedings of the 3rd International Conference on Bioinfor-matics and Biomedical Engineering,2009,1-4
[2]Ramana R B V, Radhika M M, Subbaiah K V. Texture classification method using wavelet transforms based on Gaussian Markov random field. International Journal of Signal and Image Processing,2010,1(1):35-39
[3]Ali S K, Zou B J. Analysis and classification of remote sensing by using wavelet transform and neural network. International conference on computer science and software engineering,2008,963-966
[4]Hassana G K, Ali S K, Zou B J. Optimal approach for texture analysis and Clas-sification based on Wavelet Transform and Neural Network. Journal of Information Hiding and Multimedia Signal Processing,2011,2(1):33-40
[5]Peng L X, Li S L, Zhang S. Differential diagnosis in electron microscope of constitutional jaundice and chronic hepatitis.Rcenet development of electronic mi-croscopy,1990,175-177.
[6]Samih H N, Glen S M, Anthony M V, Zhimin Y, Liqun C, Vivette D D. Thin basement membrane nephropathy cannot be diagnosed reliably in deparaffinised, formalin-fixed tissue. Nephrol Dial Transplant.2007.22(4):1228-1232.
[7]Ali S K, Zou B J, Ali A K. An algorithm for diagnosis of the three kinds of con-stitutional jaundice. International arabic journal of information technology,2010,7 (4):448-455.
[8]William K P. digital image processing. Los Altos, California,2007
[9]Hanmandlu M, Devendra J, Rochak S. Color image enhancement by fuzzy inten-sification. Pattern Recognition,15th International Conference on Pattern Recogni-tion, Barcelona, Spain,2000,3(1):310-313
[10]Feng H M. A self-tuning fuzzy control system design.9th IFSA World Congress and 20th NAFIPS International Conference, Vancouver,2001,209-214
[11]Thyagarajan K S. Digital image processing with application to digital cinema. Focal Press is an imprint of Elsevier,2006
[12]Yunfeng Z, Shanshan G, Caiming Z, Jing C. Application of a bivariate rational interpolation in image zooming. International journal of innovative computing, in-formation and Control,2009,5(11):4299-4307
[13]Cristian M, Agostinho R. Gray-scale image enhancement as an automatic process driven by evolution. IEEE Transaction on Systems, MAN, AND Cybernetics Part B 2004,34(2):1292-1298
[14]Hasan O. Automatic detection of epileptic seizures in EEG using discrete wavelet transform and approximate entropy. Elsevier expert Systems with Applications 2009,36(2):2027-2036
[15]Duan M B, Chris D, Guoping Q. Tone-mapping high dynamic range images by novel histogram adjustment. Pattern Recognition, Elsevier,2010,43:1847-1862
[16]Peter S, Athanassios S, Touradj E. The jpeg 2000 suite. John Wiley and sons ltd. 2009.
[17]Menotti D. Contrast enhancement in digital imaging using histogram equalization: [PhD dissertation]. Minas Gerais:Universidade Federal de Minas Gerais,2008
[18]Yongyue Z, Michael B, Stephen S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Transactions on medical imaging,2001,20(1):45-57
[19]Qing W, Ward R, Zou J. Contrast enhancement for enlarged images based on edge sharpening. Image Processing, IEEE International Conference ICIP,2005, 2(5):762-767
[20]Zhang S, Karim M A. A new impulse detector for switching median filters. IEEE Signal Processing Letters,2002,9(11):360-363
[21]Linda G S, George C S. Computer Vision. New Jersey, Prentice-Hall,2001, 279-325
[22]Giurgiutiu V, Cuc A, Goodman P. Review of vibration-based helicopters health and usage monitoring methods.55th Meeting of the Society for Machinery Failure Prevention Technology, Virginia Beach. VA, USA,2001,267-295
[23]Mertins A. Wavelets filter banks, time-frequency transforms and applications. John Wiley & Sons Ltd,1999,196-217
[24]Chebil J, Noel G, Mesbah M, Deriche M. Wavelet decomposition for the detection and diagnosis of faults in rolling element bearings. Jordan Journal of Mechanical and Industrial Engineering,2009,3 (4):260-267
[25]Vimal Krishnan V R, Babu Anto P. Features of wavelet packet decomposition and discrete wavelet transform for malayalam speech recognition. International Journal of Recent Trends in Engineering,2009,1(2):93-96
[26]Chik Z, Islam T, Rosyidi, S A, Sanusi H, Taha M R, Mustafa, M M. Comparing the performance of fourier decomposition and wavelet decomposition for seismic signal analysis. European Journal of Scientific Research,2009,32:(3),314-328
[27]Niyazi K, Pelin G, Osman N U, Ahmet S. Mammographic mass detection using wavelets as input to neural networks. Journal of Medical Systems, Springer Neth-erlands,2009,34 (6):1083-1088
[28]Chi-Man P. Rotation-invariant texture feature for image retrieval. Computer Vi-sion and Image Understanding,2003,89(1):24-43
[29]Sadashivappa G, AnandaBabu K V S. Evaluation of wavelet filters for image compression. World Academy of Science, Engineering and Technology,2009,51 (1):131-137
[30]Linda G S, George C S. Computer Vision New Jersey, Prentice-Hall,2001, 279-325
[31]Popovici A, Popovici D. Cellular automata in image processing. In Fifteenth In-ternational Symposium on Mathematical Theory of Networks and System,2002,1-6
[32]Hernandez G, Herrmann H J. Cellular automata for elementary image enhance-ment, CVGIP:Graphical Model and Image Processing,1996,58(1):82-89
[33]Sindeyev M, Konushin V. A novel interactive image matting framework. Graphics and Media Lab, Moscow State Lomonosov University,2008,41-45
[34]Chuang Y, Curless B, Salesin D, Szeliski R. A Bayesian approach to digital mat-ting". Proc. of IEEE CVPR,2001,264-271
[35]Sindeyev M, Konushin V, Vezhnevets V. Improvements of Bayesian matting. Proc. of Graphicon, MSU Graphics and Media Lab. Computer Science department of Moscow State University,2007,23-27
[36]Boykov Y, JollyM P. Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images. In Proc. of the International Conference on Computer Vision,2001,1(2):105-112
[37]Holalu S S, Arumugam K. Breast tissue classification using statistical feature extraction of mammograms. Medical Imaging and Information Sciences,2006,23 (3):105-107
[38]Srinivasan G N, Shobha G. Statistical texture analysis. Proceedings of science, engineering and technology,2008,36(3):1264-1269
[39]Vamsi K M, Prasad Y. An in depth comparison of four texture segmentation methods. Digital Image Computing Techniques and Applications, IEEE,2007, 366-372
[40]Thawar A, Zyad S, Lala K, Sami B. Object Classification Via Geometrical, Zernike and Legendre Moments. Journal of Theoretical and Applied Information Technology,2009,7(1):31-37
[41]Jin Kyu P, EenJun H, Yunyoung N. Utilizing venation features for efficient leaf image retrieval. The Journal of Systems and Software,2008,81(1):71-82
[42]Abby A, Goodrum A. Image information retrieval:An Overview of Current Re-search. Special issues on current research,2000,3(2):1-5
[43]Zadeh L A. Fuzzy sets. Information and Control,1965,8(3):338-353
[44]Sushmita M, Sankar K P. Fuzzy sets in pattern recognition and machine intelli-gence. Fuzzy Sets and Systems, Elsevier,2005,156(3):381-386
[45]Tavakkolil M, Jamali A, Ebrahimi A. New method to evaluate financial per-formance of companies by fuzzy logic:Case Study Drug I Industry of Iran. Asia Pacific Journal of Finance and Banking Research,2010,4(4):10-24
[46]Babak R. Desulfurization process using Takagi-Sugeno-Kang fuzzy modeling. International journal, advanced manufacturing technology, Springer,2010,46(1): 191-197
[47]Simone A, Ludwig V P, Andriy H. Fuzzy Approach for the Evaluation of Trust and Reputation of Services. International Conference on Fuzzy Systems,2009,2(1): 115-120
[48]Sukran Y, Esra Y. A two stage fuzzy model for domestic wastewater treatment plant control. Water Observation and Information Systems conference BALWOIS, 2010,33(7):1-9
[49]Feng H M. Hybrid stages particle swarm optimization learning fuzzy modeling systems design. Tamkang Journal of Science and Engineering,2006,9(2):167-176
[50]Fan C, Li K K, Chan W L, Yu W,Zhang Z. Application of wavelet fuzzy neural network in locating single line to ground fault (SLG) in distribution lines. Elsevier, Electrical Power and Energy Systems,2007,29(6):497-503
[51]Markus K. Interactive Image Retrieval Using Self-Organization Maps. [Ph.D Dissertation]. Finland:Helsinki University of Technology,2003
[52]Umugwaneza M P. A novel similarity measure for trademark images retrieval based on genetic algorithm.[MSc thesis]. Changsha:Central South University,2009
[53]Smeulders A, Worring M., Santini S, Gupta A, Jain R. Content-based image re-trieval at the end of the ealy years. IEEE Trans. Pattern Anal. Mach. Intell,2002, 22(12):1349-1380
[54]I Wei J, Guihua E, Qiong H, Jinwei G D. Similarity-based online feature selection in content-based image retrieval. IEEE transactions on image processing,2006, 15(3):702-712
[55]Ricardo da S T,Alexandre X F. Content-based image retrieval. Theory and Ap-plications. RITA,2006,13(2):1-21
[56]Hichem B, Lobna H, Bechir A. Survey of the adequate descriptor for content-based image retrieval on the web:global versus local features. CORIA Conference en Recherched'Information et Applications,2009,445-456
[57]Flickner M. Querybyimage and video content:the QBIC system" IEEE Comput. 1995,28(9):23-32
[58]Pentland A, Picard R W, Sclaro S. Photobook:tools for content-based manipula-tion of image databases. Proceedings of the SPIE Storage and Retrieval for Image and Video Databases,1994,2185(1):34-47
[59]Smith J R, Chang S F. VisualSEEk:a full automated content-based image querysy stem, Proceedings of the Fourth ACM International Multimedia Conference,1996, 87-98
[60]Bach J R. The virage image search engine:an open framework for image man-agement. SPIE Digital Image Storage Arch. System,1996,2670(1):76-87
[61]Minka T. An image database browser that learns from user interaction.[MSc the-sis]. Boston:Massachusetts Institute of Technology,1996
[62]Minka T, Picard R. Interactive learning with a society of models. Pattern Recog-nition in image data base,1997,30(4):565-582
[63]Ma W Y, Manjunath B S. NeTra. a tool book for navigating large image databases. Proceedings of the ICIP'97, Santa Barbara, CA,1997,1 (1):568-571
[64]Kellyet P M. Query by image example:the CANDID approach. Proceedings of the SPIE Storage and Retrieval of Image and Video Databases,1995,238-248
[65]Liang K C, Jay Kuo C C. Wave Guide:a joint wavelet-based image representation and description system. IEEE Trans. Image Process,1999,8(11):619-1629
[66]Hirata K, Kato T. Query by visual example. Proceedings of the Third International Conference Extending Database Technology:Advances in Database Technology EDBT,1992,92(1):56-71
[67]Gonzalez R C, Woods R E. Digital Image Processing, Addison-Wesley. Reading, MA,1992.
[68]Huang P W, Da S K. Image retrieval by texture similarity. Pattern Recognition. 2003,36(1):665-679
[69]Eng H M. A self-tuning fuzzy control system design.9th IFSA World Congress and 20th NAFIPS International Conference, Vancouver,2001,209-214
[70]Hu M K. Visual pattern recognition by moment invariants. IRE Trans. Inform. Theory IT-8,1962,179-187
[71]Taubin G, Cooper D.B. Recognition and positioning of rigid objects using alge-braic moment invariants. SPIE Conference on Geometric Methods in Computer Vision,1991,1570(1):175-186
[72]Teh C.-H, Chin R T. On image analysis by the methods of moments. IEEE Trans. Pattern Anal. Mach. Intell,1988,10(4):496-513
[73]Liao S X, Pawlak M. On image analysis by moments. IEEE Trans. Pattern Anal. Mach. Intell,1996,18(3):254-266
[74]Teague M R. Image analysis via the general theory of moments. Journal of the Optical Society of America,1980,70(8):920-930
[75]Niblack W. The QBIC project:querying images by content using color, texture and shape. SPIE Conference on Storage and Retrieval for Image and Video Databases, 1993,1908(1):173-187
[76]Lu G J, Sajjanhar A. Region-based shape representation and similarity measure suitable for content-based image retrieval. Multimedia System,1999,7(1):165-174
[77]Safar M, Shahabi C, Sun X. Image retrieval by shape:a comparative study. IEEE International Conference on Multimedia and Expo, New York, USA,2000,1-5
[78]Chakrabarti K, Binderberger M O, Porkaev K, Mehrotra S. Similar shape retrieval in MARS. Proc. IEEE Int. Conf. on Multimedia and Expo, New York, USA,2000, 2(1):709-712
[79]Paquet E, Rioux M, Murching A, Naveen T, Tabatabai A. Description of shape information for 2-D and 3-D objects, Signal Processing. Image Communication, 2000,16(1):103-122
[80]Dengsheng Z, Lu G. Shape-based image retrieval using generic Fourier descriptor Signal Processing. Image Communication,2002,17(1):825-848
[81]Nastar, C. Towards applying content-based image retrieval in the clinical routine. Future Generation Computer Systems,2007,23(3):466-474
[82]AlGarni G, Hamiane M. A novel technique for automatic shoeprint image re-trieval". Forensic Science International,2008,181(1):10-14
[83]Hafiane A, Chaudhuri S, Seetharaman G, Zavidovique B. Region-based CBIR in GIS with local space filling curves to spatial representation. Pattern Recognition Letters,2006,27(4):259-267
[84]Muller H, Michoux N, Bandon D, Geissbuhler A. A review of content-based image retrieval systems in medical applications-clinical benefits and future direc-tions. International Journal of Medical Informatics,2004,73(2):1-23
[85]Rahman M M, Bhattacharya P, Desai B C. A framework for medical image re-trieval using machine learning and statistical similarity matching techniques with relevance feedback. IEEE Transactions on Information Technology in Biomedicine, 2007,11(1):58-69
[86]Horsthemke W, Raicu D, Furst J. Content-based Image Retrieval for Biomedical Image Archives:Achievements, Problems, and Prospects. A MICCAI 2007 Work-shop,2007,31-44
[87]Greenspan H, Pinhas A T. Medical image categorization and retrieval for PACS using the GMM-KL framework. IEEE Transactions on Information Technology in Biomedicine,2007,11 (2):190-202
[88]Kim J, Cai W, Feng D, Wu H. A new way for multidimensional medical data management:volume of interest (VOI)-based retrieval of medical images with visual and functional features. IEEE Transactions on Information Technology in Biomedicine,2006,10(3):598-607
[89]Doyle S, Hwang M, Naik S, Feldman M, Tomaszeweski J, Madabhushi A. Using manifold learning for content-based image retrieval of prostate histopathology. Workshop on Content-based Image Retrieval for Biomedical Image Archives,2007, 53-62
[90]Balmashnova E, Platel B, Florack L M J, Ter Haar R, Contentbased B M. Image retrieval by means of scale-space top-points and differential invariants. Workshop on Content-based Image Retrieval for Biomedical Image Archives,2007,83-92
[91]Quellec G, Lamard M, Cazuguel G, Cochener B, Roux C. Wavelet optimization for content-based image retrieval in medical databases. Medical Image Analysis, El-sevier,2010,14(2):227-241
[92]Zhiming C, Guangming Z. A novel medical image dynamic fuzzy classification model based on Ridgelet transform. Journal of software,2010,5(5):458-465
[93]Ramirez J, Gorriz J M, Segovia F, Chaves R, Salas-Gonzalez D, Lopez M, Alvarez I, Padilla P. Computer aided diagnosis system for the Alzheimer's disease based on partial least squares and random forest SPECT image classification. Neuroscience Letters, Elsevier Ireland Ltd,2010,472(2):99-103
[94]Sean P, Roche M D. Jaundice in the adult patient. American Academy of Family Physicians,2004,69(2):299-304
[95]Feng H M. Hybrid stages particle swarm optimization learning fuzzy modeling systems design. Tamkang Journal of Science and Engineering,2006,9:(2),167-176
[96]Jianguo Z, Tieniu T. Brief of invariant texture analysis method. Pattern Recogni-tion,2002,35(1):735-747
[97]Jayabharata R M, Mohanta D K. A wavelet-fuzzy combined approach for classi-fication and location of transmission line faults. Elsevier, Electrical Power and En-ergy Systems,2007,29(1):669-678
[98]Jayabharata M R, Mohanta D K. A wavelet-fuzzy combined approach for clas-sification and location of transmission line faults. Elsevier, Electrical Power and Energy Systems,2007,29(1):669-678
[99]Shortliffe E H. Computer-Based Medical Consultations. MYCIN, Elsevier, North-Holland, New York.1976
[100]Adlassnig K P. Fuzzy set theory in medical diagnosis. IEEE Trans. Syst. Man Cybernet,1986,16(3):260-265
[101]Fieschi M, Joubert M, Fieschi D, Roux M. SPHINX-a system for com-puter-aided diagnosis. Methods Inform. Med,1982,21(3):143-148
[102]Belmonte-Serrano M, Sierra C, Lopezde-Mantaras R R. an expert system us-ing fuzzy logic for rheumatology diagnosis. International Journal. Intell. Syst.1994. 9(1):985-1000
[103]Kohout L J, Bandler W, Anderson J, Trayner C. A knowledge-based decision support system for use in medicine. Proceedings of UNICOM Seminar, North-Holland, Amsterdam, the Netherlands,1986,133-146.
[104]Pilz U, Engelmann L. Integration of medical knowledge in an expert system for use in intensive care medicine. H.N. Teodorescu. A. Kandel, L.C. Jain (Eds.), Fuzzy and Neuro-Fuzzy Systems in Medicine, CRC Press, Boca Raton, FL, USA, 1999.
[105]Soula G, Vialettes B, San Marco J L. A fuzzy deduction rule system:applica-tion to the treatment of diabetes. Proceedings of MEDINFO 83, Amsterdam, The Netherlands,1983,233-255
[106]Isaka S, Sebald A V. An adaptive fuzzy controller for blood pressure regula-tion, images of the twenty-first century. Proceedings of Annual International Con-ference of the IEEE Engineering in Medicine and Biology Society, New York, USA, 1989,1763-1764
[107]Kovarlerchuk B, Triantaphyllou E, Ruiz J F, Clayton J. Fuzzy logic in com-puter-aided breast cancer diagnosis, analysis of lobulation, Artif. Intell. Med,1997, 11(2):75-85
[108]Sohani M, Makki B, Sadati N, Kermani K K, Sadati A. A Neuro-Fuzzy Ap-proach to Diagnosis of Neonatal Jaundice" Bio-Inspired Models of Network, In-formation and Computing Systems,2006,1-4
[109]Ozyilmaz L, Yildirim T. Artificial neural networks for diagnosis of hepatitis disease. Proceedings of the Int. Joint Conf. on NN,2003,1:(1),586-589
[110]Roychowdhury A, Pratihar D K, Bose N, Sankaranarayanan K P, Sudhahar N. Diagnosis of the diseases using a GA-fuzzy approach. Elsevier Information Sci-ences,2004,162(2):105-120
[111]Donna L H, Maurice E C. Hybrid system approach to medical diagnosis. In-ternational Symposium on virmal Environments. Hum-Computer Interfaces and measurement Systems Lugano, Switzerland,2003,90-93
[112]Savige K R, Tonna S, Buzza M, Dagher H, WangY. Thin basement membrane nephropathy. Kidney International journal,2003,64(2):1169-1178
[113]Karl T, Jaakko P. Thin basement membrane nephropathy. Journal of the American Society of Nephrology,2006,17(3):813-822
[114]Constantinos D. Thin basement membrane nephropathy:is there genetic pre-disposition to more severe disease?. Pediatr Nephrol Springer.2009.877-879.
[115]Pourghassem H, Ghassemian H. Content-based medical image classification using a new hierarchical merging scheme. Computerized Medical Imaging and Graphics,2008,32(8):651-661
[116]Roth K S, Amaker B H, Chan J C M. Pediatric hematuria and thin basement membrane nephropathy:What is it and what does it mean?. Clin Pediatr,2001, 40(1):607-613
[117]Lanteri M, Wilson D, Savige J. Clinical features in 2 patients with IgA glomerulonephritis and thin basement membrane disease. Nephrol Dial Transplant journal,1996,11(5):791-793
[118]Nieuwhof C M G, de Heer F, de Leeuw P, van Breda Vriesman P J C. ThinGBMnephropathy:Premature glomerular obsolescence is associated with hy-pertension and late onset renal failure. Kidney International journal,1997,51(1): 1596-1601
[119]Ioanna Z,Sophia A, Andreas H, Thalia M, Mary V, Charalambos P, Alkis P, Kyriacos I, Kyriacos K. Incidence of Thin Basement Membrane Nephropathy in 990 Consecutive Renal Biopsies Examined with Electron Microscopy. Ultrastructural Pathology,2008,32(6):221-226
[120]Liew A W C, Lowand N F, Nguyen D T. Direct Reconstruction Method for Wavelet Transform Extreme Representation. IEE Processing- Vis. Image signal processing,1997,144(4):193-198
[121]Unser, M.. "Texture Classification and Segmentation Using Wavelet Frames". IEEE transaction on image processing,1995.4(11:1549-1560
[122]Daubechies I. What do wavelet come from personal point of view. Process-ing of IEEE,1996,84(4):510-513
[123]Samih H N, Glen S M, Anthony M. V, Zhimin Y, Liqun C, Vivette D D. Thin basement membrane nephropathy cannot be diagnosed reliably in deparaffinized, formalin-fixed tissue. Nephrol Dial Transplant,2007,22(4):1228-1232
[124]Sidney Burrus. Introduction to wavelets and wavelet transforms. Prentice Hall,1998
[125]Feacher I, Jevis B W. Digital Signal Processing, Practical Approach. Addi-son-Wasley com.1996
[126]Mala K, Sadasivam V, Alagappan S. Neural network based texture analysis of liver tumor from computed tomography images. International Journal of Biological and Medical Sciences,2007,1(1):33-37
[127]Watson G H, Watson S K. Wavelet transform on vector spaces as a method of multiscale image characterization. IEEE proceeding on Image processing,1997, 144(2):89-97
[128]Knight K. Connectionist Ideas and Algorithms. Communication of the ACM, 1990,33(11):58-74
[129]Wilson R. Multiresolution Image Modeling. Electronic and communication Journal,1997,9(2):90-96
[130]Negia L S H, Sjoberg J, Viberg M. Adaptive Neural Nets Filter Using Re-cursive Levenberg-Marquardt search Describtion". Signals, Systems & Computers, 1998,4(1):1-4
[131]Vezhnevets V, Konouchine V. GrowCut:Interactive multi-label N-D image segmentation by cellular automata. Proc. of Graphicon,2005,150-156
[132]Isaacn B. Handbook of medical imaging processing and analysis. Academic press, Harcourt science and technology company,2000
[133]Peters B O, Pfurtscheller G, Flyvbjerg H. Automatic differentiation of mul-tichannel EEG signals. IEEE Trans. Biomed. Eng,2001,48(1):111-116
[134]Gopinath P, Reddy N P. Toward intelligent web monitoring:performance of committee neural networks vs. single neural network. Proceedings of the IEEE In-ternational Conference on Information Technology Applications in Biomedicine, 2000,179-182
[135]Al Mahtab M, Fazal Karim M, Salimur R, Abul Barkat M A. Dubin-Johnson syndrome with systemic lupus erythematosus:a case report. Hepatobiliary Pancreat Dis Int,2006,5(4):617-619
[136]Jan V J. Electron Microscopy in human medicine. The liver McGraw-Hill, 1979
[137]Jeong-Seon P, Myung-Joo O, Soonhee H. Fish disease diagnosis system based on image processing of pathogens'microscopic images. Frontiers in the Conver-gence of Bioscience and Information Technologies,2007,878-883
[138]Savige J, Rana K, Tonna S, Buzza M, Dagher H, Wang Y. Thin basement membrane nephropathy. Kidney International,2003,64(3):1169-1178
[139]Paul S T. Alport syndrome and thin basement membrane nephropathy. Nephron Clin Pract,2007,106(4):82-88
[140]Steven W S. The scientist and engineer's guide to digital signal processing. California Technical Publishing, second edition,1999
[141]Jon D D. Convex Optimization & Euclidean Distance Geometry. Meboo Publishing USA,2005
[142]Lee H J, Ahn H J, Song J H, Park R H. Hough transform for line and plane detection based on the conjugate formulation. Machine Vision Applications in Industrial Inspection,2001,4301(2):244-25
[2]Ramana R B V, Radhika M M, Subbaiah K V. Texture classification method using wavelet transforms based on Gaussian Markov random field. International Journal of Signal and Image Processing,2010,1(1):35-39
[3]Ali S K, Zou B J. Analysis and classification of remote sensing by using wavelet transform and neural network. International conference on computer science and software engineering,2008,963-966
[4]Hassana G K, Ali S K, Zou B J. Optimal approach for texture analysis and Clas-sification based on Wavelet Transform and Neural Network. Journal of Information Hiding and Multimedia Signal Processing,2011,2(1):33-40
[5]Peng L X, Li S L, Zhang S. Differential diagnosis in electron microscope of constitutional jaundice and chronic hepatitis.Rcenet development of electronic mi-croscopy,1990,175-177.
[6]Samih H N, Glen S M, Anthony M V, Zhimin Y, Liqun C, Vivette D D. Thin basement membrane nephropathy cannot be diagnosed reliably in deparaffinised, formalin-fixed tissue. Nephrol Dial Transplant.2007.22(4):1228-1232.
[7]Ali S K, Zou B J, Ali A K. An algorithm for diagnosis of the three kinds of con-stitutional jaundice. International arabic journal of information technology,2010,7 (4):448-455.
[8]William K P. digital image processing. Los Altos, California,2007
[9]Hanmandlu M, Devendra J, Rochak S. Color image enhancement by fuzzy inten-sification. Pattern Recognition,15th International Conference on Pattern Recogni-tion, Barcelona, Spain,2000,3(1):310-313
[10]Feng H M. A self-tuning fuzzy control system design.9th IFSA World Congress and 20th NAFIPS International Conference, Vancouver,2001,209-214
[11]Thyagarajan K S. Digital image processing with application to digital cinema. Focal Press is an imprint of Elsevier,2006
[12]Yunfeng Z, Shanshan G, Caiming Z, Jing C. Application of a bivariate rational interpolation in image zooming. International journal of innovative computing, in-formation and Control,2009,5(11):4299-4307
[13]Cristian M, Agostinho R. Gray-scale image enhancement as an automatic process driven by evolution. IEEE Transaction on Systems, MAN, AND Cybernetics Part B 2004,34(2):1292-1298
[14]Hasan O. Automatic detection of epileptic seizures in EEG using discrete wavelet transform and approximate entropy. Elsevier expert Systems with Applications 2009,36(2):2027-2036
[15]Duan M B, Chris D, Guoping Q. Tone-mapping high dynamic range images by novel histogram adjustment. Pattern Recognition, Elsevier,2010,43:1847-1862
[16]Peter S, Athanassios S, Touradj E. The jpeg 2000 suite. John Wiley and sons ltd. 2009.
[17]Menotti D. Contrast enhancement in digital imaging using histogram equalization: [PhD dissertation]. Minas Gerais:Universidade Federal de Minas Gerais,2008
[18]Yongyue Z, Michael B, Stephen S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Transactions on medical imaging,2001,20(1):45-57
[19]Qing W, Ward R, Zou J. Contrast enhancement for enlarged images based on edge sharpening. Image Processing, IEEE International Conference ICIP,2005, 2(5):762-767
[20]Zhang S, Karim M A. A new impulse detector for switching median filters. IEEE Signal Processing Letters,2002,9(11):360-363
[21]Linda G S, George C S. Computer Vision. New Jersey, Prentice-Hall,2001, 279-325
[22]Giurgiutiu V, Cuc A, Goodman P. Review of vibration-based helicopters health and usage monitoring methods.55th Meeting of the Society for Machinery Failure Prevention Technology, Virginia Beach. VA, USA,2001,267-295
[23]Mertins A. Wavelets filter banks, time-frequency transforms and applications. John Wiley & Sons Ltd,1999,196-217
[24]Chebil J, Noel G, Mesbah M, Deriche M. Wavelet decomposition for the detection and diagnosis of faults in rolling element bearings. Jordan Journal of Mechanical and Industrial Engineering,2009,3 (4):260-267
[25]Vimal Krishnan V R, Babu Anto P. Features of wavelet packet decomposition and discrete wavelet transform for malayalam speech recognition. International Journal of Recent Trends in Engineering,2009,1(2):93-96
[26]Chik Z, Islam T, Rosyidi, S A, Sanusi H, Taha M R, Mustafa, M M. Comparing the performance of fourier decomposition and wavelet decomposition for seismic signal analysis. European Journal of Scientific Research,2009,32:(3),314-328
[27]Niyazi K, Pelin G, Osman N U, Ahmet S. Mammographic mass detection using wavelets as input to neural networks. Journal of Medical Systems, Springer Neth-erlands,2009,34 (6):1083-1088
[28]Chi-Man P. Rotation-invariant texture feature for image retrieval. Computer Vi-sion and Image Understanding,2003,89(1):24-43
[29]Sadashivappa G, AnandaBabu K V S. Evaluation of wavelet filters for image compression. World Academy of Science, Engineering and Technology,2009,51 (1):131-137
[30]Linda G S, George C S. Computer Vision New Jersey, Prentice-Hall,2001, 279-325
[31]Popovici A, Popovici D. Cellular automata in image processing. In Fifteenth In-ternational Symposium on Mathematical Theory of Networks and System,2002,1-6
[32]Hernandez G, Herrmann H J. Cellular automata for elementary image enhance-ment, CVGIP:Graphical Model and Image Processing,1996,58(1):82-89
[33]Sindeyev M, Konushin V. A novel interactive image matting framework. Graphics and Media Lab, Moscow State Lomonosov University,2008,41-45
[34]Chuang Y, Curless B, Salesin D, Szeliski R. A Bayesian approach to digital mat-ting". Proc. of IEEE CVPR,2001,264-271
[35]Sindeyev M, Konushin V, Vezhnevets V. Improvements of Bayesian matting. Proc. of Graphicon, MSU Graphics and Media Lab. Computer Science department of Moscow State University,2007,23-27
[36]Boykov Y, JollyM P. Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images. In Proc. of the International Conference on Computer Vision,2001,1(2):105-112
[37]Holalu S S, Arumugam K. Breast tissue classification using statistical feature extraction of mammograms. Medical Imaging and Information Sciences,2006,23 (3):105-107
[38]Srinivasan G N, Shobha G. Statistical texture analysis. Proceedings of science, engineering and technology,2008,36(3):1264-1269
[39]Vamsi K M, Prasad Y. An in depth comparison of four texture segmentation methods. Digital Image Computing Techniques and Applications, IEEE,2007, 366-372
[40]Thawar A, Zyad S, Lala K, Sami B. Object Classification Via Geometrical, Zernike and Legendre Moments. Journal of Theoretical and Applied Information Technology,2009,7(1):31-37
[41]Jin Kyu P, EenJun H, Yunyoung N. Utilizing venation features for efficient leaf image retrieval. The Journal of Systems and Software,2008,81(1):71-82
[42]Abby A, Goodrum A. Image information retrieval:An Overview of Current Re-search. Special issues on current research,2000,3(2):1-5
[43]Zadeh L A. Fuzzy sets. Information and Control,1965,8(3):338-353
[44]Sushmita M, Sankar K P. Fuzzy sets in pattern recognition and machine intelli-gence. Fuzzy Sets and Systems, Elsevier,2005,156(3):381-386
[45]Tavakkolil M, Jamali A, Ebrahimi A. New method to evaluate financial per-formance of companies by fuzzy logic:Case Study Drug I Industry of Iran. Asia Pacific Journal of Finance and Banking Research,2010,4(4):10-24
[46]Babak R. Desulfurization process using Takagi-Sugeno-Kang fuzzy modeling. International journal, advanced manufacturing technology, Springer,2010,46(1): 191-197
[47]Simone A, Ludwig V P, Andriy H. Fuzzy Approach for the Evaluation of Trust and Reputation of Services. International Conference on Fuzzy Systems,2009,2(1): 115-120
[48]Sukran Y, Esra Y. A two stage fuzzy model for domestic wastewater treatment plant control. Water Observation and Information Systems conference BALWOIS, 2010,33(7):1-9
[49]Feng H M. Hybrid stages particle swarm optimization learning fuzzy modeling systems design. Tamkang Journal of Science and Engineering,2006,9(2):167-176
[50]Fan C, Li K K, Chan W L, Yu W,Zhang Z. Application of wavelet fuzzy neural network in locating single line to ground fault (SLG) in distribution lines. Elsevier, Electrical Power and Energy Systems,2007,29(6):497-503
[51]Markus K. Interactive Image Retrieval Using Self-Organization Maps. [Ph.D Dissertation]. Finland:Helsinki University of Technology,2003
[52]Umugwaneza M P. A novel similarity measure for trademark images retrieval based on genetic algorithm.[MSc thesis]. Changsha:Central South University,2009
[53]Smeulders A, Worring M., Santini S, Gupta A, Jain R. Content-based image re-trieval at the end of the ealy years. IEEE Trans. Pattern Anal. Mach. Intell,2002, 22(12):1349-1380
[54]I Wei J, Guihua E, Qiong H, Jinwei G D. Similarity-based online feature selection in content-based image retrieval. IEEE transactions on image processing,2006, 15(3):702-712
[55]Ricardo da S T,Alexandre X F. Content-based image retrieval. Theory and Ap-plications. RITA,2006,13(2):1-21
[56]Hichem B, Lobna H, Bechir A. Survey of the adequate descriptor for content-based image retrieval on the web:global versus local features. CORIA Conference en Recherched'Information et Applications,2009,445-456
[57]Flickner M. Querybyimage and video content:the QBIC system" IEEE Comput. 1995,28(9):23-32
[58]Pentland A, Picard R W, Sclaro S. Photobook:tools for content-based manipula-tion of image databases. Proceedings of the SPIE Storage and Retrieval for Image and Video Databases,1994,2185(1):34-47
[59]Smith J R, Chang S F. VisualSEEk:a full automated content-based image querysy stem, Proceedings of the Fourth ACM International Multimedia Conference,1996, 87-98
[60]Bach J R. The virage image search engine:an open framework for image man-agement. SPIE Digital Image Storage Arch. System,1996,2670(1):76-87
[61]Minka T. An image database browser that learns from user interaction.[MSc the-sis]. Boston:Massachusetts Institute of Technology,1996
[62]Minka T, Picard R. Interactive learning with a society of models. Pattern Recog-nition in image data base,1997,30(4):565-582
[63]Ma W Y, Manjunath B S. NeTra. a tool book for navigating large image databases. Proceedings of the ICIP'97, Santa Barbara, CA,1997,1 (1):568-571
[64]Kellyet P M. Query by image example:the CANDID approach. Proceedings of the SPIE Storage and Retrieval of Image and Video Databases,1995,238-248
[65]Liang K C, Jay Kuo C C. Wave Guide:a joint wavelet-based image representation and description system. IEEE Trans. Image Process,1999,8(11):619-1629
[66]Hirata K, Kato T. Query by visual example. Proceedings of the Third International Conference Extending Database Technology:Advances in Database Technology EDBT,1992,92(1):56-71
[67]Gonzalez R C, Woods R E. Digital Image Processing, Addison-Wesley. Reading, MA,1992.
[68]Huang P W, Da S K. Image retrieval by texture similarity. Pattern Recognition. 2003,36(1):665-679
[69]Eng H M. A self-tuning fuzzy control system design.9th IFSA World Congress and 20th NAFIPS International Conference, Vancouver,2001,209-214
[70]Hu M K. Visual pattern recognition by moment invariants. IRE Trans. Inform. Theory IT-8,1962,179-187
[71]Taubin G, Cooper D.B. Recognition and positioning of rigid objects using alge-braic moment invariants. SPIE Conference on Geometric Methods in Computer Vision,1991,1570(1):175-186
[72]Teh C.-H, Chin R T. On image analysis by the methods of moments. IEEE Trans. Pattern Anal. Mach. Intell,1988,10(4):496-513
[73]Liao S X, Pawlak M. On image analysis by moments. IEEE Trans. Pattern Anal. Mach. Intell,1996,18(3):254-266
[74]Teague M R. Image analysis via the general theory of moments. Journal of the Optical Society of America,1980,70(8):920-930
[75]Niblack W. The QBIC project:querying images by content using color, texture and shape. SPIE Conference on Storage and Retrieval for Image and Video Databases, 1993,1908(1):173-187
[76]Lu G J, Sajjanhar A. Region-based shape representation and similarity measure suitable for content-based image retrieval. Multimedia System,1999,7(1):165-174
[77]Safar M, Shahabi C, Sun X. Image retrieval by shape:a comparative study. IEEE International Conference on Multimedia and Expo, New York, USA,2000,1-5
[78]Chakrabarti K, Binderberger M O, Porkaev K, Mehrotra S. Similar shape retrieval in MARS. Proc. IEEE Int. Conf. on Multimedia and Expo, New York, USA,2000, 2(1):709-712
[79]Paquet E, Rioux M, Murching A, Naveen T, Tabatabai A. Description of shape information for 2-D and 3-D objects, Signal Processing. Image Communication, 2000,16(1):103-122
[80]Dengsheng Z, Lu G. Shape-based image retrieval using generic Fourier descriptor Signal Processing. Image Communication,2002,17(1):825-848
[81]Nastar, C. Towards applying content-based image retrieval in the clinical routine. Future Generation Computer Systems,2007,23(3):466-474
[82]AlGarni G, Hamiane M. A novel technique for automatic shoeprint image re-trieval". Forensic Science International,2008,181(1):10-14
[83]Hafiane A, Chaudhuri S, Seetharaman G, Zavidovique B. Region-based CBIR in GIS with local space filling curves to spatial representation. Pattern Recognition Letters,2006,27(4):259-267
[84]Muller H, Michoux N, Bandon D, Geissbuhler A. A review of content-based image retrieval systems in medical applications-clinical benefits and future direc-tions. International Journal of Medical Informatics,2004,73(2):1-23
[85]Rahman M M, Bhattacharya P, Desai B C. A framework for medical image re-trieval using machine learning and statistical similarity matching techniques with relevance feedback. IEEE Transactions on Information Technology in Biomedicine, 2007,11(1):58-69
[86]Horsthemke W, Raicu D, Furst J. Content-based Image Retrieval for Biomedical Image Archives:Achievements, Problems, and Prospects. A MICCAI 2007 Work-shop,2007,31-44
[87]Greenspan H, Pinhas A T. Medical image categorization and retrieval for PACS using the GMM-KL framework. IEEE Transactions on Information Technology in Biomedicine,2007,11 (2):190-202
[88]Kim J, Cai W, Feng D, Wu H. A new way for multidimensional medical data management:volume of interest (VOI)-based retrieval of medical images with visual and functional features. IEEE Transactions on Information Technology in Biomedicine,2006,10(3):598-607
[89]Doyle S, Hwang M, Naik S, Feldman M, Tomaszeweski J, Madabhushi A. Using manifold learning for content-based image retrieval of prostate histopathology. Workshop on Content-based Image Retrieval for Biomedical Image Archives,2007, 53-62
[90]Balmashnova E, Platel B, Florack L M J, Ter Haar R, Contentbased B M. Image retrieval by means of scale-space top-points and differential invariants. Workshop on Content-based Image Retrieval for Biomedical Image Archives,2007,83-92
[91]Quellec G, Lamard M, Cazuguel G, Cochener B, Roux C. Wavelet optimization for content-based image retrieval in medical databases. Medical Image Analysis, El-sevier,2010,14(2):227-241
[92]Zhiming C, Guangming Z. A novel medical image dynamic fuzzy classification model based on Ridgelet transform. Journal of software,2010,5(5):458-465
[93]Ramirez J, Gorriz J M, Segovia F, Chaves R, Salas-Gonzalez D, Lopez M, Alvarez I, Padilla P. Computer aided diagnosis system for the Alzheimer's disease based on partial least squares and random forest SPECT image classification. Neuroscience Letters, Elsevier Ireland Ltd,2010,472(2):99-103
[94]Sean P, Roche M D. Jaundice in the adult patient. American Academy of Family Physicians,2004,69(2):299-304
[95]Feng H M. Hybrid stages particle swarm optimization learning fuzzy modeling systems design. Tamkang Journal of Science and Engineering,2006,9:(2),167-176
[96]Jianguo Z, Tieniu T. Brief of invariant texture analysis method. Pattern Recogni-tion,2002,35(1):735-747
[97]Jayabharata R M, Mohanta D K. A wavelet-fuzzy combined approach for classi-fication and location of transmission line faults. Elsevier, Electrical Power and En-ergy Systems,2007,29(1):669-678
[98]Jayabharata M R, Mohanta D K. A wavelet-fuzzy combined approach for clas-sification and location of transmission line faults. Elsevier, Electrical Power and Energy Systems,2007,29(1):669-678
[99]Shortliffe E H. Computer-Based Medical Consultations. MYCIN, Elsevier, North-Holland, New York.1976
[100]Adlassnig K P. Fuzzy set theory in medical diagnosis. IEEE Trans. Syst. Man Cybernet,1986,16(3):260-265
[101]Fieschi M, Joubert M, Fieschi D, Roux M. SPHINX-a system for com-puter-aided diagnosis. Methods Inform. Med,1982,21(3):143-148
[102]Belmonte-Serrano M, Sierra C, Lopezde-Mantaras R R. an expert system us-ing fuzzy logic for rheumatology diagnosis. International Journal. Intell. Syst.1994. 9(1):985-1000
[103]Kohout L J, Bandler W, Anderson J, Trayner C. A knowledge-based decision support system for use in medicine. Proceedings of UNICOM Seminar, North-Holland, Amsterdam, the Netherlands,1986,133-146.
[104]Pilz U, Engelmann L. Integration of medical knowledge in an expert system for use in intensive care medicine. H.N. Teodorescu. A. Kandel, L.C. Jain (Eds.), Fuzzy and Neuro-Fuzzy Systems in Medicine, CRC Press, Boca Raton, FL, USA, 1999.
[105]Soula G, Vialettes B, San Marco J L. A fuzzy deduction rule system:applica-tion to the treatment of diabetes. Proceedings of MEDINFO 83, Amsterdam, The Netherlands,1983,233-255
[106]Isaka S, Sebald A V. An adaptive fuzzy controller for blood pressure regula-tion, images of the twenty-first century. Proceedings of Annual International Con-ference of the IEEE Engineering in Medicine and Biology Society, New York, USA, 1989,1763-1764
[107]Kovarlerchuk B, Triantaphyllou E, Ruiz J F, Clayton J. Fuzzy logic in com-puter-aided breast cancer diagnosis, analysis of lobulation, Artif. Intell. Med,1997, 11(2):75-85
[108]Sohani M, Makki B, Sadati N, Kermani K K, Sadati A. A Neuro-Fuzzy Ap-proach to Diagnosis of Neonatal Jaundice" Bio-Inspired Models of Network, In-formation and Computing Systems,2006,1-4
[109]Ozyilmaz L, Yildirim T. Artificial neural networks for diagnosis of hepatitis disease. Proceedings of the Int. Joint Conf. on NN,2003,1:(1),586-589
[110]Roychowdhury A, Pratihar D K, Bose N, Sankaranarayanan K P, Sudhahar N. Diagnosis of the diseases using a GA-fuzzy approach. Elsevier Information Sci-ences,2004,162(2):105-120
[111]Donna L H, Maurice E C. Hybrid system approach to medical diagnosis. In-ternational Symposium on virmal Environments. Hum-Computer Interfaces and measurement Systems Lugano, Switzerland,2003,90-93
[112]Savige K R, Tonna S, Buzza M, Dagher H, WangY. Thin basement membrane nephropathy. Kidney International journal,2003,64(2):1169-1178
[113]Karl T, Jaakko P. Thin basement membrane nephropathy. Journal of the American Society of Nephrology,2006,17(3):813-822
[114]Constantinos D. Thin basement membrane nephropathy:is there genetic pre-disposition to more severe disease?. Pediatr Nephrol Springer.2009.877-879.
[115]Pourghassem H, Ghassemian H. Content-based medical image classification using a new hierarchical merging scheme. Computerized Medical Imaging and Graphics,2008,32(8):651-661
[116]Roth K S, Amaker B H, Chan J C M. Pediatric hematuria and thin basement membrane nephropathy:What is it and what does it mean?. Clin Pediatr,2001, 40(1):607-613
[117]Lanteri M, Wilson D, Savige J. Clinical features in 2 patients with IgA glomerulonephritis and thin basement membrane disease. Nephrol Dial Transplant journal,1996,11(5):791-793
[118]Nieuwhof C M G, de Heer F, de Leeuw P, van Breda Vriesman P J C. ThinGBMnephropathy:Premature glomerular obsolescence is associated with hy-pertension and late onset renal failure. Kidney International journal,1997,51(1): 1596-1601
[119]Ioanna Z,Sophia A, Andreas H, Thalia M, Mary V, Charalambos P, Alkis P, Kyriacos I, Kyriacos K. Incidence of Thin Basement Membrane Nephropathy in 990 Consecutive Renal Biopsies Examined with Electron Microscopy. Ultrastructural Pathology,2008,32(6):221-226
[120]Liew A W C, Lowand N F, Nguyen D T. Direct Reconstruction Method for Wavelet Transform Extreme Representation. IEE Processing- Vis. Image signal processing,1997,144(4):193-198
[121]Unser, M.. "Texture Classification and Segmentation Using Wavelet Frames". IEEE transaction on image processing,1995.4(11:1549-1560
[122]Daubechies I. What do wavelet come from personal point of view. Process-ing of IEEE,1996,84(4):510-513
[123]Samih H N, Glen S M, Anthony M. V, Zhimin Y, Liqun C, Vivette D D. Thin basement membrane nephropathy cannot be diagnosed reliably in deparaffinized, formalin-fixed tissue. Nephrol Dial Transplant,2007,22(4):1228-1232
[124]Sidney Burrus. Introduction to wavelets and wavelet transforms. Prentice Hall,1998
[125]Feacher I, Jevis B W. Digital Signal Processing, Practical Approach. Addi-son-Wasley com.1996
[126]Mala K, Sadasivam V, Alagappan S. Neural network based texture analysis of liver tumor from computed tomography images. International Journal of Biological and Medical Sciences,2007,1(1):33-37
[127]Watson G H, Watson S K. Wavelet transform on vector spaces as a method of multiscale image characterization. IEEE proceeding on Image processing,1997, 144(2):89-97
[128]Knight K. Connectionist Ideas and Algorithms. Communication of the ACM, 1990,33(11):58-74
[129]Wilson R. Multiresolution Image Modeling. Electronic and communication Journal,1997,9(2):90-96
[130]Negia L S H, Sjoberg J, Viberg M. Adaptive Neural Nets Filter Using Re-cursive Levenberg-Marquardt search Describtion". Signals, Systems & Computers, 1998,4(1):1-4
[131]Vezhnevets V, Konouchine V. GrowCut:Interactive multi-label N-D image segmentation by cellular automata. Proc. of Graphicon,2005,150-156
[132]Isaacn B. Handbook of medical imaging processing and analysis. Academic press, Harcourt science and technology company,2000
[133]Peters B O, Pfurtscheller G, Flyvbjerg H. Automatic differentiation of mul-tichannel EEG signals. IEEE Trans. Biomed. Eng,2001,48(1):111-116
[134]Gopinath P, Reddy N P. Toward intelligent web monitoring:performance of committee neural networks vs. single neural network. Proceedings of the IEEE In-ternational Conference on Information Technology Applications in Biomedicine, 2000,179-182
[135]Al Mahtab M, Fazal Karim M, Salimur R, Abul Barkat M A. Dubin-Johnson syndrome with systemic lupus erythematosus:a case report. Hepatobiliary Pancreat Dis Int,2006,5(4):617-619
[136]Jan V J. Electron Microscopy in human medicine. The liver McGraw-Hill, 1979
[137]Jeong-Seon P, Myung-Joo O, Soonhee H. Fish disease diagnosis system based on image processing of pathogens'microscopic images. Frontiers in the Conver-gence of Bioscience and Information Technologies,2007,878-883
[138]Savige J, Rana K, Tonna S, Buzza M, Dagher H, Wang Y. Thin basement membrane nephropathy. Kidney International,2003,64(3):1169-1178
[139]Paul S T. Alport syndrome and thin basement membrane nephropathy. Nephron Clin Pract,2007,106(4):82-88
[140]Steven W S. The scientist and engineer's guide to digital signal processing. California Technical Publishing, second edition,1999
[141]Jon D D. Convex Optimization & Euclidean Distance Geometry. Meboo Publishing USA,2005
[142]Lee H J, Ahn H J, Song J H, Park R H. Hough transform for line and plane detection based on the conjugate formulation. Machine Vision Applications in Industrial Inspection,2001,4301(2):244-25