Computerized analysis of mammographic microcalcifications in morphological and texture feature spaces

We are developing computerized feature extraction and classification methods to analyze malignant and benign microcalcifications on digitized mammograms. Morphological features that described the size, contrast, and shape of microcalcifications and their variations within a cluster were designed to characterize microcalcifications segmented from the mammographic background. Texture features were derived from the spatial gray-level dependence (SGLD) matrices constructed at multiple distances and directions from tissue regions containing microcalcifications. A genetic algorithm (GA) based feature selection technique was used to select the best feature subset from the multi-dimensional feature spaces. The GA-based method was compared to the commonly used feature selection method based on the stepwise linear discriminant analysis (LDA) procedure. Linear discriminant classifiers using the selected features as input predictor variables were formulated for the classification task. The discriminant scores output from the classifiers were analyzed by receiver operating characteristic (ROC) methodology and the classification accuracy was quantified by the area, Az, under the ROC curve. We analyzed a data set of 145 mammographic microcalcification clusters in this study. It was found that the feature subsets selected by the GA-based method are comparable to or slightly better than those selected by the stepwise LDA method. The texture features (Az = 0.84) were more effective than morphological features (Az = 0.79) in distinguishing malignant and benign microcalcifications. The highest classification accuracy (Az = 0.89) was obtained in the combined texture and morphological feature space. The improvement was statistically significant in comparison to classification in either the morphological (p = 0.002) or the texture (p = 0.04) feature space alone. The classifier using the best feature subset from the combined feature space and an appropriate decision threshold could correctly identify 35% of the benign clusters without missing a malignant cluster. When the average discriminant score from all views of the same cluster was used for classification, the Az value increased to 0.93 and the classifier could identify 50% of the benign clusters at 100% sensitivity for malignancy. Alternatively, if the minimum discriminant score from all views of the same cluster was used, the Az value would be 0.90 and a specificity of 32% would be obtained at 100% sensitivity. The results of this study indicate the potential of using combined morphological and texture features for computer-aided classification of microcalcifications.     

Comparison of classification approaches applied to NIR-spectra of clinical study lots

NIR-spectroscopy combined with pattern recognition approaches is applied to classify samples of clinical study lots in the pharmaceutical industry. The performance of linear discriminant analysis (LDA), quadratic discriminant analysis (QDA) and K-nearest neighbour (KNN) method is evaluated on a tablet data set and a capsule data set. To establish a classification model a strategy is followed, which is described in this work. Frequently, in the pharmaceutical industry, several batches of the same clinical study lot are produced. We tested whether it is possible to merge several batches in one class for modelling or, instead, whether it is necessary to model each batch individually.     

Issues in the use and interpretation of discriminant analysis.

Reviews the 2 problems for which a discriminant analysis is used—separation and classification. Issues related to the use and interpretation of a discriminant analysis are those pertaining to (a) distinguishing between a linear discriminant function and a linear classification function, (b) misusing stepwise discriminant analysis programs, (c) ordering variables and selecting variable subsets, (d) choosing a classification rule, (e) estimating true classification hit rates, (f) assessing classification accuracy, and (g) examining and using classification results. Most of these issues deal with information available from package discriminant analysis computer programs. (50 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

核函数方法(上)

支撑矢量机的成功引起了人们对核函数方法的兴趣。通过某种非线性变换将输入空间映射到一个高维特征空间，如果在其中应用标准的线性算法时，其分量间的相互作用仅限于内积，则可以利用函数的技术将这种算法转换为原输入空间里的非线性算法。Fisher判别法和主分析法是在模式分类与特征抽取中已经获得广泛应用的传统线性方法，近年出现的基于核函数的Fisher判别（KFD）与基于核函数的主分量分析（KPCA）是它们的非线性推广，其性能更好，适用范围更广，灵活性更高，是值得关注的应用前景看好的新技术。     

Towards a method for automated classification of 1H MRS spectra from brain tumours

Recent studies have shown that MRS can substantially improve the non-invasive categorization of human brain tumours. However, in order for MRS to be used routinely by clinicians, it will be necessary to develop reliable automated classification methods that can be fully validated. This paper is in two parts: the first part reviews the progress that has been made towards this goal, together with the problems that are involved in the design of automated methods to process and classify the spectra. The second part describes the development of a simple prototype system for classifying 1H single voxel spectra, obtained at an echo time (TE) of 135 ms, of the four most common types of brain tumour (meningioma (MM), astrocytic (AST), oligodendroglioma (OD) and metastasis (ME)) and cysts. This system was developed in two stages: firstly, an initial database of spectra was used to develop a prototype classifier, based on a linear discriminant analysis (LDA) of selected data points. Secondly, this classifier was tested on an independent test set of 15 newly acquired spectra, and the system was refined on the basis of these results. The system correctly classified all the non-astrocytic tumours. However, the results for the the astrocytic group were poorer (between 55 and 100%, depending on the binary comparison). Approximately 50% of high grade astrocytoma (glioblastoma) spectra in our data base showed very little lipid signal, which may account for the poorer results for this class. Consequently, for the refined system, the astrocytomas were subdivided into two subgroups for comparison against other tumour classes: those with high lipid content and those without.     

Supervised mutual-information based feature selection for motor unit action potential classification

A new supervised mutual information-based feature selection method is presented. Using real motor unit action potential (MUAP) data from 10 EMG signals, the performances of 32 time-sample feature sets, feature subsets selected using first- and second-order mutual information and features obtained using linear discriminant analysis (LDA) and principal component analysis (PCA) were evaluated using a minimum Euclidean distance (MED) classifier. The evaluation showed that by using only 20 first-order features or only 15 second-order features mean error rates and error rate variations equivalent to using all 32 samples or LDA or PCA could be obtained. The computational cost of first-order feature selection was considerably less than LDA, PCA and second-order feature selection. The performance of first-order features was further evaluated using a more robust classifier. Unlike the MED classifier, the robust classifier only assigned a candidate MUAP if the assignment was sufficiently certain. For the robust classifier the average error rates using 20 features were similar to using the full feature set, yet higher assignment rates were obtained. Results from both evaluations suggest that the sets of first-order features were an efficient representation of lower dimension, which provided high accuracy classification with reduced computational requirements.     

Assessment of hyperactivity: Isolating measures of high discriminant ability.

Research into hyperactivity has produced many psychometric and observational instruments that demonstrate deficits in hyperactive children vis-à-vis their normal peers. Despite the considerable information that such instruments have yielded, their relative power in classifying a child as hyperactive has yet to be explored. The present study focused on this issue by employing a discriminant analysis with a number of instruments, some of which have been used extensively by researchers and clinicians. Results indicate that only 9 of the 27 measures used had the potential to discriminate 13 hyperactive from 13 normal males (mean age 7 yrs 2 mo). Of these, only 3 measures were needed in the discriminant equation for accurate classification. It is suggested that isolating the best discriminant measures will improve clinical assessment and will be central to basic and applied research on hyperactivity. (36 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Classification of 1H MR spectra of biopsies from untreated and recurrent ovarian cancer using linear discriminant analysis

Proton (1H) magnetic resonance (MR) spectra of ex vivo biopsy samples of ovarian cancers provided biochemical information that was used to discriminate cancer from normal ovarian tissue. Possible differences present in intrinsically resistant tumors or changes in biochemistry after the induction of resistance were identified. Using multivariate techniques, in particular linear discriminant analysis (LDA), ovarian cancer was distinguished from normal ovarian tissue with a sensitivity of 100%, a specificity of 95% and an accuracy of 98%. Moreover, LDA was able to distinguish untreated ovarian cancer from recurrent ovarian cancer with a sensitivity of 92%, a specificity of 100%, and an accuracy of 97%; removal of the single "fuzzy" specimen increased the accuracy to 100%. Applications of this knowledge to in vivo measurements could lead to noninvasive diagnosis of ovarian cancer.     

Prediction of screening decisions in a medical school admission process

To study and predict a set of screening decisions in a medical school admission process, a stratified sample (n = 864) was drawn to represent the range of applicants to the Yale University School of Medicine during a single year. A smaller sample from the following year's applicants was drawn in a similar fashion for purposes of cross-validation. Prior to the prediction of screening decisions, a set of independent variables was selected by a factor analytic procedure from credentials in an applicant's admission folder. These folder variables ranged in nature from quantitative measures of academic performance to demographic information and types of extracurricular activity. Two multivariate statistical procedures, Sonquist's Automatic Interaction Detection (AID) and linear discriminant analysis (LDA), were used to predict screening decisions. Measures of academic performance and ability proved to be the most effective predictors of screening decision, as evidenced in the final AID tree and the discriminant function.     

What to do with significant multivariate effects in multivariate analyses of variance.

Significant multivariate tests in the multivariate analysis of variance are often followed by analyses of the contributions of individual dependent variables to those significant effects. There has been little agreement, however, as to which specific analyses should be performed. The use of the 2 most common techniques, analysis of variance for each dependent variable and discriminant analysis, are discussed and then illustrated in a computer study. It is suggested that the purpose of the user should determine the technique chosen as the 2 methods are not alternative approaches to the same problem. Analysis of variance can be used for hypothesis testing of individual variables and is appropriate for research. The value of discriminant analysis is in prediction and classification, although it can indicate complex relationships between measures in hypothesis testing. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Computer-assisted image classification: use of neural networks in anatomic pathology

Artificial neural networks (ANN) implemented on digital computers have received much attention for interpretation of images in pathology and cytology. Most such images are too complex for current ANN to interpret directly; instead, ANN usually classify the images according to numeric features extracted from them. In experiments on three distinct image classification problems, ANN classifiers performed as well or better than multivariate linear discriminant analysis (a traditional parametric statistical classifier). ANN empirically define non-linear multivariate decision boundaries, and can combine non-contiguous feature areas in mapping a classification. However, many training cases are required in order to map complex area boundaries precisely and with a low risk of 'overtraining.' Careful problem selection and attention to data dimensionality and format are important for efficient ANN use.     

Caregiver burden at 6 months following severe traumatic brain injury

We compared the performance of three computer based classification methods when applied to the problem of detecting microaneurysms on digitised angiographic images of the retina. An automated image processing system segmented 'candidate' objects (microaneurysms or spurious objects), and produced a list of features on each candidate for use by the classifiers. We compared an empirically derived rule based system with two automated methods, linear discriminant analysis and a learning vector quantiser artificial neural network, to classify the objects as microaneurysms or otherwise. ROC analysis shows that the rule based system gave a higher performance than the other methods (p = 0.92) although a much greater development time is required.     

Use of discriminant analysis in counseling psychology research.

Discriminant analysis is a technique for the multivariate study of group differences. More specifically, it provides a method of examining the extent to which multiple predictor variables are related to a categorical criterion, that is, group membership. Situations in which the technique is particularly useful include those in which the researcher wishes to assess which of a number of continuous variables best differentiates groups of individuals or in which he or she wishes to predict group membership on the basis of the discriminant function (analogous to a multiple regression equation) yielded by the analysis. The method is also useful as a follow-up to a significant analysis of variance. In this article, I describe the method of discriminant analysis, including the concept of discriminant function, discriminant score, group centroid, and discriminant weights and loadings. I discuss methods for testing the statistical significance of a function, methods of using the function in classification, and the concept of rotation functions. The use of discriminant analysis in both the two-group case and the multigroup case is illustrated. Finally, I provide a number of illustrative examples of use of the method in the counseling literature. I conclude with cautions regarding the use of the method and with the provision of resources for further study. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A review of data-derived methods for assigning causes of death from verbal autopsy data

BACKGROUND: Verbal autopsy (VA) is an indirect method for estimating cause-specific mortality. In most previous studies, cause of death has been assigned from verbal autopsy data using expert algorithms or by physician review. Both of these methods may have poor validity. In addition, physician review is time consuming and has to be carried out by doctors. A range of methods exist for deriving classification rules from data. Such rules are quick and simple to apply and in many situations perform as well as experts. METHODS: This paper has two aims. First, it considers the advantages and disadvantages of the three main methods for deriving classification rules empirically; (a) linear and other discriminant techniques, (b) probability density estimation and (c) decision trees and rule-based methods. Second, it reviews the factors which need to be taken into account when choosing a classification method for assigning cause of death from VA data. RESULTS: Four main factors influence the choice of classification method: (a) the purpose for which a classifier is being developed, (b) the number of validated causes of death assigned to each case, (c) the characteristics of the VA data and (d) the need for a classifier to be comprehensible. When the objective is to estimate mortality from a single cause of death, logistic regression should be used. When the objective is to determine patterns of mortality, the choice of method will depend on the above factors in ways which are elaborated in the paper. CONCLUSION: Choice of classification method for assigning cause of death needs to be considered when designing a VA validation study. Comparison of the performance of classifiers derived using different methods requires a large VA dataset, which is not currently available.     

Assessment of statistical and neural networks methods in NMR spectral classification and metabolite selection

Magnetic resonance spectroscopy opens a window into the biochemistry of living tissue. However, spectra acquired from different tissue types in vivo or in vitro and from body fluids contain a large number of peaks from a range of metabolites, whose relative intensities vary substantially and in complicated ways even between successive samples from the same category. The realization of the full clinical potential of NMR spectroscopy relies, in part, on our ability to interpret and quantify the role of individual metabolites in characterizing specific tissue and tissue conditions. This paper addresses the problem of tissue classification by analysing NMR spectra using statistical and neural network methods. It assesses the performance of classification models from a range of statistical methods and compares them with the performance of artificial neural network models. The paper also assesses the consistency of the models in selecting, directly from the spectra, the subsets of metabolites most relevant for differentiating between tissue types. The analysis techniques are examined using in vitro spectra from eight classes of normal tissue and tumours obtained from rats. We show that, for the given data set, the performance of linear and non-linear methods is comparable, possibly due to the small sample size per class. We also show that using a subset of metabolites selected by linear discriminant analysis for further analysis by neural networks improves the classification accuracy, and reduces the number of metabolites necessary for correct classification.     

基于时-频分析的步态模式自动分类

针对不同路况和运动模式下的高维、非线性、强耦合和高时变下肢加速度信号的识别问题,提出了一种基于时——频分析的步态模式自动分类方案.利用三轴加速度传感器采集运动时小腿在矢状面、冠状面和横切面的加速度信号,利用五阶Daubechies小波基对其进行特征提取,并采用线性判别式分析进行降维,最后利用决策树和支持向量机对得到的精简步态特征进行模式分类.实验结果显示两种分类器的总体分类准确率均达到90%以上,个别步态分类可达到100%,验证了特征提取和降维方法的合理性和有效性.      

Classification of observational data with artificial neural networks versus discriminant analysis in pharmacoepidemiological studies--can outcome of fluoxetine treatment be predicted?

For several years, there has been an ongoing discussion about appropriate methodological tools to be applied to observational data in pharmacoepidemiological studies. It is now suggested by our research group that artificial neural networks (ANN) might be advantageous in some cases for classification purposes when compared with discriminant analysis. This is due to their inherent capability to detect complex linear and nonlinear functions in multivariate data sets, the possibility of including data on different scales in the same model, as well as their relative resistance to "noisy" input. In this paper, a short introduction is given to the basics of neural networks and possible applications. For demonstration, a comparison between artificial neural networks and discriminant analysis was performed on a multivariate data set, consisting of observational data of 19738 patients treated with fluoxetine. It was tested, which of the two statistical tools outperforms the two other in regard to the therapeutic response prediction from the clinical input data. Essentially, it was found that neither discriminant analysis nor ANN are able to predict the clinical outcome on the basis of the employed clinical variables. Applying ANN, we were able to rule out the possibility of undetected suppressor effects to a greater extent than would have been possible by the exclusive application of discriminant analysis.     

Identification of species of Brucella using Fourier transform infrared spectroscopy

Fourier transform infrared spectroscopy (FTIR) is a technique that has been used over the years in chemical analysis for the identification of substances and is one that may be applied to the characterisation of microorganisms. The marked tendency of Brucella towards variation in the smooth rough phase, together with the laboriousness and risk involved in the methods used in their identification, make their classification difficult. We studied the type strains of the different species and biovars of Brucella and 11 isolates of human origin of Brucella melitensis, six corresponding to biovar 1, one to biovar 2 and five to biovar 3. The results of linear discriminant analysis performed using the data provide an above 95% likelihood of correct classification, over half of which are in fact above 99% for the vast majority of Brucella strains. Only one case of B. melitensis biovar 1 has been incorrectly classified. The rest of the microorganisms studied (Staphylococcus aureus, Strteptococcus pyogenes, Enterococcus faecalis, Corynebacterium pseudodiphtheriticum, Clostridium perfringens, Escherichia coli, Acinetobacter calcoaceticus and Pseudomonas aeruginosa) have been classified correctly in all cases to a likelihood of over 80%. In the graphic representation of the analysis, a grouping of these can be seen in clusters, which include the different species. One of these comprises B. melitensis, another Brucella abortus, and another wider one is made up of Brucella suis. The Brucella canis, Brucella ovis and Brucella neotomae strains appear separate from the previously described groups.     

新疆岩金矿地质环境影响评价及防治措施

针对新疆地区岩金类矿山开采导致的地质环境问题,采用Fisher判别分析方法对矿山开采后地质环境影响程度进行分类评价。选取地质灾害、含水层破坏、地形地貌景观破坏、土地资源破坏和地质环境治理难易程度等5项指标作为判别因子,以新疆地区5家岩金类矿山的37个评价单元作为样本,建立了新疆岩金类矿山地质环境影响程度评价的Fisher判别模型,利用回代估计方法对建模数据进行逐一检验,正确率达100%。利用该模型对新疆萨尔托海金矿进行矿山地质环境影响程度判别分析,分类结果与实际分类结果完全一致。研究结果表明,Fisher判别分析法用于新疆岩金类矿山地质环境影响评价效果良好。在分类结果的基础上,提出了相应的防治措施建议。