基于高场不对称波形离子迁移谱技术的细菌培养液检测

目的:利用高场不对称波形离子迁移谱(FAIMS)技术检测并识别常见伤口感染细菌。方法:在合适的实验条件下,用FAIMS分析仪检测大肠杆菌(ECO)、金黄色葡萄球菌(SAU)、铜绿假单胞菌(PAE)的TH肉汤培养液及纯TH肉汤培养液4种样品,得到其离子迁移谱图。分析谱图形状,提取峰值点,通过对比分析确定3种细菌TH培养液的特征离子峰。结果:ECO与SAU均有有别于另外3种样品的稳定特征离子峰,特征峰可用场强、补偿电压、离子电流值3个参量确定。PAE出现两个特征峰,其中一个不稳定出现,但依然可凭借第二个特征峰进行辨别。结论:FAIMS技术可对ECO、SAU、PAE这3种常见细菌的TH肉汤细菌培养液进行甄别,为伤口感染的快速检测提供一定的可行性。     

FAIMS在爆炸伤伤口样本检测中的初步实验研究

目的:探索高场不对称离子迁移谱（FAIMS）用于伤口样本检测的可行性。方法:收集3 位爆炸伤病人5 个伤口样 本（坏死肌肉、死皮或者脓液）。每次获得2个样本,一个样本送检验科进行细菌检测;对另一个样本进行FAIMS的气味分 析,获取伤口样本的气味检测的正模式和负模式FAIMS数据。结果:检验科细菌检测结果显示5个样本中,1例是蜡样芽 孢杆菌,1 例是少见嗜铜菌,1 例是肺炎克雷伯杆菌和阴沟肠杆菌,另2 例是无细菌生长。有细菌感染的样本的FAIMS检 测的正模式和负模式图谱出现了一些特征峰,且不同细菌感染样本的FAIMS特征峰的形状与位置不同。无细菌生长样 本的FAIMS则无特征峰出现。结论:FAIMS能在一定程度反映伤口样本的感染情况,且不同的细菌感染样本的FAIMS 有所不同,提示FAIMS可用于区分伤口感染的类型。FAIMS是潜在的快速检测伤口感染的技术,值得进一步的实验研 究。     

基于四类运动想象任务的脑电信号识别

背景：脑-机接口是在大脑与外部设备之间建立的直接的交流通路,基于运动想象的脑-机接口研究已经从两类运动想象任务的识别发展到多类任务的识别。 目的：探寻准确有效的对多任务运动想象脑电信号进行特征提取及模式识别的方法。 方法：首先采用公共平均参考法减小多通道中各导联间的相关性,提高脑电信号的信噪比。并对公共空间模式算法进行扩展,采用“一对多”的策略,对4类任务的脑电信号进行特征提取,在模式识别过程中,采用基于决策树法的支持向量机进行分类。对于实验对象样本不充足,结合支持向量机和贝叶斯分类器,将分类结果中具有大概率的测试样本扩充到训练集,最后再次运用支持向量机进行分类。 结果与结论：最佳正确率达到92.78%,“一对多”的公共空间模式和基于决策树的支持向量机可以有效地进行多任务脑电信号识别,扩充样本可以提高分类正确率。     

电子鼻检测常见伤口感染细菌实验研究

为了进一步探讨电子鼻用于伤口细菌感染快速筛查的可能性,使用自制电子鼻检测鲍曼不动杆菌、大肠杆菌、肺炎克雷伯杆菌、金黄色葡萄球菌、铜绿假单胞菌的巯基乙酸酯（TH）培养液及纯培养液,经过预处理后,使用支持向量机、BP神经网络、逻辑回归3种算法进行分类。结果显示,使用BP算法,总体识别率可达93%以上。对于单个识别率,检测大肠杆菌、肺炎克雷伯杆菌、金黄色葡萄球菌可达98%以上,检测鲍曼不动杆菌、铜绿假单胞菌可达90%左右,表明电子鼻用于伤口感染细菌的快速检测和识别具有一定的可行性。     

生物阻抗身份识别研究

本研究旨在根据生物电阻抗理论和模式识别算法,对不同的人进行生物电阻抗的测量并进行身份识别。本文利用AD5933芯片设计阻抗采集电路来采集不同人手掌处的生物电阻抗,获得1~100kHz频率激励下的阻抗谱信息,计算阻抗曲线的分段频率值作为特征参数。为了提高识别率和防止过度拟合,将被测人员分成训练集和测试集,设计了一个3层的向后传播(BP)神经网络模型,对样本进行训练和预测。研究结果表明,BP神经网络对测试样本能进行有效识别,训练集的准确率达到97.62%,验证集的准确率达到88.79%,测试集的准确率达到86.34%,综合的识别准确率为94.22%。该网络可以很好地识别出已存在于训练网络中的人和不属于训练网络的陌生人,验证了基于生物电阻抗的模式识别方法对身份进行辨识的可行性与可靠性,为身份识别提供了一个简单有效的补充性技术。     

支持向量机后验概率方法在多任务脑机接口中的应用

基于运动想象的脑机接口是把使用者的运动意图转变成控制外部设备的信号,它包括脑电数据采集,特征提取和模式分类等几个基本环节.本研究发展了用支持向量机后验概率输出进行分类,并用分类结果中具有大概率的测试样本扩充训练集的模式分类与特征更新方法,并把此方法应用于4类任务运动想象脑机接口实验.使用BCI Competition Ⅲ的数据Ⅲa,运用一对一共空间模式扩展方法进行特征提取,用支持向量机后验概率方法进行分类和训练样本扩充.结果表明:概率信息能提高BCI的性能;应用概率信息选取样本扩充训练集能增加分类器的稳健性.     

脑电信号的小波特征提取及半监督识别方法的研究

特征提取和特征分类是脑机接口中模式识别过程中两个关键的环节。首先,针对脑电信号的非平稳特性,提出基于小波变换系数、系数均值及小波熵相结合的特征提取方法,该方法在特征中加入了脑电信号的能量信息。实验证明,通过该方法获取的P300信号特征量能够更好地表达脑电信号中的瞬变成分,进一步提高了识别率;其次,在模式识别方面,改进了基于自训练半监督的支持向量机算法和基于自训练半监督的K均值聚类算法。在BCI2003竞赛数据集上的实验表明,相比于传统的BP神经网络,两种改进的分类算法在获得了更高的识别率的同时,能够将特征量维数降低一个数量级,明显提高了训练收敛速度,有效增加了基于脑电信号的实时BCI系统的可实现性。     

一种用于鉴别体域网动作模式的近邻快速鲁棒协作表示分类算法

提出一种近邻快速鲁棒协作表示体域网动作模式分类算法,旨在基于体域网多传感动作模式数据结构内在相似性,利用最近邻原则,寻找与测试动作样本密切相关的少量近邻类别和近邻训练样本,重新构造训练样本集,然后基于新训练样本集构建快速鲁棒协作表示动作分类模型,通过扩展拉格朗日乘数算法求解待定测试样本协作表示系数和表示残差,定义判定测试样本所属类别规则,有效提高分类性能。采用公开的美国加州伯克利大学多传感动作模式数据库验证所提算法有效性。结果表明,所提算法能够从体域网多传感数据中获得更多与动作模式密切相关的协调性和相关性,动作模式识别率提高2%,运行时间仅需6.5 s,分类性能明显优于稀疏表示动作模式分类性能,有望为临床鉴别人体动作模式提供一个新的技术解决方案。     

面向帕金森病语音数据挖掘的分包融合集成算法

基于语音数据挖掘实现帕金森病诊断的方法近年来已被证明有效。然而,受数据采集对象患病程度以及采集设备和环境等因素影响,所获取数据集的样本空间中存在不同类别样本混叠现象。混叠区域的样本难以有效识别,严重影响了算法的分类准确度。为了解决这一问题,本文提出了分包融合集成算法,通过设计类心距离比值来衡量样本的混叠程度并将训练集划分成多个子集,再利用错误分类样本传递式训练的方法调整子集划分结果,最后通过优化子分类权重对各个子分类器的测试结果进行加权融合。实验结果表明,本文方法分类准确度在两个公共数据集上都得到明显提高,平均准确度最大提高可达25.44%。该方法不仅有效提高了帕金森病语音数据集分类准确度,还增加了样本利用率,为帕金森病语音诊断提供了一种新思路。     

核共空域子空间分解特征提取算法研究

脑-机接口中特征提取算法是脑电信号处理的关键步骤。提出一种基于核方法的核共空域子空间分解特征提取算法,将用于多通道两类别分类的共空域子空间分解算法推广到核空间。应用新算法对BCI竞赛Ⅱ的数据集Ⅳ进行实验仿真。实验中核函数使用的是线性核函数,求解空域滤波器时,为了减小计算的压力,在原空间对每一个试验的训练数据进行层次聚类,训练的分类器为最近邻分类器,实验的测试集结果为84%,与数据集Ⅳ的竞赛胜利者的分类结果相同。     

基于随机投影的快速稀疏表示人体动作识别方法

为有效解决体域网人体行为动作远程识别系统低功耗和快速准确识别的问题,提出一种基于随机投影的快速稀疏表示人体行为动作识别的方法。该方法基于压缩感知随机投影方式压缩数据,获取待测试样本邻近类中较少最近邻训练样本,构建测试样本稀疏表示时的训练样本集,以期达到最优线性重构测试样本;在降低传感器装置功耗和稀疏表示识别算法计算复杂度基础上,捕捉人体行为动作本质特征信息,提高多类别动作识别率。采用国际公开可穿戴传感器动作识别数据库WARD多类别动作数据,验证所提算法的有效性。实验结果表明,当数据压缩率为50%,所提算法能够获得最高平均识别率（92.78%）,比传统稀疏表示分类算法获得的动作识别率提高近5%,并显著降低其相应的运行时间,能准确稀疏表示多类别人体行为动作信号,有效降低稀疏表示分类算法的计算法复杂度和运行时间,明显提高多类别动作识别率,为构建快速稀疏表示动作识别提供一个新的思路和方法。     

Testing by Sysmex UF-100 flow cytometer and with bacterial culture in a diagnostic laboratory: a comparison

A large proportion of the samples tested in routine diagnostic microbiology laboratory are urine samples. The gold standard is bacterial culture, but a high proportion of samples cultured are negative. Unnecessary testing can be reduced and an improved service provided by an effective screening test. The Sysmex UF-100 flow cytometer has been developed to count cells and casts accurately in urine samples. Its performance in a screening test was compared with bacterial culture by using 1005 consecutive urine samples, and cut-off criteria were established. Cut-off values of 3000 bacteria/microl and 111 WBC/microl provided the best discrimination. Of 1005 samples, 606 (60%) would be cultured. Sixteen samples that were not selected according to these criteria were culture positive. This was considered acceptable for our routine use. The use of a testing algorithm incorporating the Sysmex UF-100 flow cytometer has improved the quality and efficiency of urine testing within the routine microbiology laboratory.     

Eukaryotic Promoter Recognition Using Back propagation Neural Network

A new system is developed to recognize promoter sequences from non-promoter sequences based on position weight matrix and backpropagation neural network in this paper. The system performs significantly better on the training set and the test set, the mean recognition rate is as high as 99% on the training set and 97% on the testing set. Experimental results demonstrate the effectiveness of the system to recognize the promoter sequences that have been trained and the promoter sequences that have not been seen previously.     

Spectral classifier design with ensemble classifiers and misclassification-rejection: application to elastic-scattering spectroscopy for detection of colonic neoplasia

Optical spectroscopy has shown potential as a real-time, in vivo, diagnostic tool for identifying neoplasia during endoscopy. We present the development of a diagnostic algorithm to classify elastic-scattering spectroscopy (ESS) spectra as either neoplastic or non-neoplastic. The algorithm is based on pattern recognition methods, including ensemble classifiers, in which members of the ensemble are trained on different regions of the ESS spectrum, and misclassification-rejection, where the algorithm identifies and refrains from classifying samples that are at higher risk of being misclassified. These "rejected" samples can be reexamined by simply repositioning the probe to obtain additional optical readings or ultimately by sending the polyp for histopathological assessment, as per standard practice. Prospective validation using separate training and testing sets result in a baseline performance of sensitivity = .83, specificity = .79, using the standard framework of feature extraction (principal component analysis) followed by classification (with linear support vector machines). With the developed algorithm, performance improves to Se ～ 0.90, Sp ～ 0.90, at a cost of rejecting 20-33% of the samples. These results are on par with a panel of expert pathologists. For colonoscopic prevention of colorectal cancer, our system could reduce biopsy risk and cost, obviate retrieval of non-neoplastic polyps, decrease procedure time, and improve assessment of cancer risk.     

Rapid detection of Mycobacterium tuberculosis in contaminated BACTEC 12B broth cultures by testing with Amplified Mycobacterium Tuberculosis Direct Test

Contamination of broth cultures of acid-fast bacilli (AFB) by bacterial species other than Mycobacterium species frequently occurs. Many of these contaminated cultures require redecontamination and reincubation before the appropriate tests can be performed for identification, significantly affecting the turnaround time for reporting culture results. In this study, the Amplified Mycobacterium Tuberculosis Direct Test (MTD; Gen-Probe) was performed to detect the Mycobacterium tuberculosis complex (MTBC) in 125 BACTEC 12B broth cultures with positive growth indices. Among these, 41 grew non-AFB bacteria only, and all 41 were negative by the MTD. The remaining 84 bottles contained contaminated cultures that grew both AFB and other bacteria or yeasts. Repeat decontamination and reincubation of these specimens required a mean time of 13 days (range, 3 to 40 days). The MTD results were positive for 10 samples, 9 of which were MTBC culture positive and 1 of which grew Myobacterium celatum, a species known to cross-react in the MTD. All cultures growing other mycobacterial species were negative by the MTD. The results of this study demonstrate that the MTD is both sensitive and specific in detecting MTBC in contaminated broth cultures and that, when used selectively, the MTD can potentially rule in or out a diagnosis of MTBC as much as 12 days earlier than using nonamplified DNA probe testing alone can.     

语音识别技术在咳嗽声音自动识别中的应用

目的 利用双门限法、Mel频率倒普系数(MFCC)法及矢量量化(VQ)法的语音识别技术对咳嗽声音进行自动识别.方法 在安静环境下,对5例健康成年人和15例咳嗽患者的非咳嗽和咳嗽声音进行录音,分别随机分为训练样本和测试样本.训练样本用于生成咳嗽识别软件的码本,并用该码本对测试样本进行自动识别分析.同时与人工识别的结果进行对比,计算敏感性、特异性,记录两种方法的识别时间.结果 用于码本生成的咳嗽声音和非咳嗽声音均为200次,测试样本的咳嗽和非咳嗽声音分别为375次和125次.人工识别和通过码本自动识别测试样本的时间分别为33 min 18 s和1min 35 s;码本自动识别咳嗽声音的敏感性和特异性分别为98.93%和100%.结论 基于VQ的双门限法及MFCC法可用于咳嗽声音的自动识别.     

中医舌色苔色识别中多分类支撑向量机算法比较

为了了解在不同样本数量情况下具有何种特点的算法比较适合舌色苔色的分类与识别,本文采用一对一SVM和它的改进算法IMP-BSVM、DAG-BSVM,以及M-ary分类SVM算法对舌色、苔色分类进行了比较。结果表明,当训练样本数较少时,DAG-BSVM能有较高的正确率和较快的训练和识别速度,因此选用DAG-BSVM;当训练样本数较大时,IMP-BSVM能得到最少的支撑向量和较高的正确率以及较快的识别速度,此时则选用IMP-BSVM。     

Serum Peptidome Profiling in Patients with Lung Cancer

Serum peptide profiling is a promising approach for classification of cancer versus noncancer samples. In this study, we aimed to search for discriminating peptide patterns in serum samples between lung cancer patients and healthy controls. The magnetic beads‐based weak cation‐exchange chromatography followed by matrix‐assisted laser desorption and ionization time‐of‐flight mass spectrometry (MALDI‐TOF MS) was used in this study to identify patients with lung cancer. In total, serum samples from 64 lung cancer patients (32 for training set and 32 for testing set), 64 healthy controls (32 for training set and 32 for testing set), and 10 COPD patients (for disease control) were analyzed in this study. The mass spectra data analyzed with ClinProTools software was used to distinguish between cancer patients and healthy individuals based on three different algorithm models (GA, SNN, and QC). In the training set, patients with lung cancer could be identified with the mean sensitivity of 98.9% and specificity of100%. Similar results could be obtained from testing set, showing 87% sensitivity and 84.8% specificity. Screening for serum peptide patterns using MALDI‐TOF MS showed high sensitivity and specificity in identifying patients with lung cancer. Anat Rec, 2010. © 2010 Wiley‐Liss, Inc.