SEARCH · PubMed Health

Results for “feature selection”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 271 records · Page 15Linked to original sources

PubMed2003

Discovering compact and highly discriminative features or feature combinations of drug activities using support vector machines.

Nowadays, high throughput experimental techniques make it feasible to examine and collect massive data at the molecular level. These data, typically mapped to a very high dimensional feature space, carry rich information about functionalities of certain chemical or biological entities and can be used to infer valuable knowledge for the purposes of classification and prediction. Typically, a small number of features or feature combinations may play determinant roles in functional discrimination. The identification of such features or feature combinations is of great importance. In this paper, we study the problem of discovering compact and highly discriminative features or feature combinations from a rich feature collection. We employ the support vector machine as the classification means and aim at finding compact feature combinations. Comparing to previous methods on feature selection, which identify features solely based on their individual roles in the classification, our method is able to identify minimal feature combinations that ultimately have determinant roles in a systematic fashion. Experimental study on drug activity data shows that our method can discover descriptors that are not necessarily significant individually but are most significant collectively.

Algorithms↗

PubMed2006

Exploring predictive and reproducible modeling with the single-subject FIAC dataset.

Predictive modeling of functional magnetic resonance imaging (fMRI) has the potential to expand the amount of information extracted and to enhance our understanding of brain systems by predicting brain states, rather than emphasizing the standard spatial mapping. Based on the block datasets of Functional Imaging Analysis Contest (FIAC) Subject 3, we demonstrate the potential and pitfalls of predictive modeling in fMRI analysis by investigating the performance of five models (linear discriminant analysis, logistic regression, linear support vector machine, Gaussian naive Bayes, and a variant) as a function of preprocessing steps and feature selection methods. We found that: (1) independent of the model, temporal detrending and feature selection assisted in building a more accurate predictive model; (2) the linear support vector machine and logistic regression often performed better than either of the Gaussian naive Bayes models in terms of the optimal prediction accuracy; and (3) the optimal prediction accuracy obtained in a feature space using principal components was typically lower than that obtained in a voxel space, given the same model and same preprocessing. We show that due to the existence of artifacts from different sources, high prediction accuracy alone does not guarantee that a classifier is learning a pattern of brain activity that might be usefully visualized, although cross-validation methods do provide fairly unbiased estimates of true prediction accuracy. The trade-off between the prediction accuracy and the reproducibility of the spatial pattern should be carefully considered in predictive modeling of fMRI. We suggest that unless the experimental goal is brain-state classification of new scans on well-defined spatial features, prediction alone should not be used as an optimization procedure in fMRI data analysis.

Artifacts↗

PubMed2004

Mining mass spectra for diagnosis and biomarker discovery of cerebral accidents.

In this paper we try to identify potential biomarkers for early stroke diagnosis using surface-enhanced laser desorption/ionization mass spectrometry coupled with analysis tools from machine learning and data mining. Data consist of 42 specimen samples, i.e., mass spectra divided in two big categories, stroke and control specimens. Among the stroke specimens two further categories exist that correspond to ischemic and hemorrhagic stroke; in this paper we limit our data analysis to discriminating between control and stroke specimens. We performed two suites of experiments. In the first one we simply applied a number of different machine learning algorithms; in the second one we have chosen the best performing algorithm as it was determined from the first phase and coupled it with a number of different feature selection methods. The reason for this was 2-fold, first to establish whether feature selection can indeed improve performance, which in our case it did not seem to confirm, but more importantly to acquire a small list of potentially interesting biomarkers. Of the different methods explored the most promising one was support vector machines which gave us high levels of sensitivity and specificity. Finally, by analyzing the models constructed by support vector machines we produced a small set of 13 features that could be used as potential biomarkers, and which exhibited good performance both in terms of sensitivity, specificity and model stability.

Adult↗

PubMed2006

Eigengene-based linear discriminant model for tumor classification using gene expression microarray data.

MOTIVATION: The nearest shrunken centroids classifier has become a popular algorithm in tumor classification problems using gene expression microarray data. Feature selection is an embedded part of the method to select top-ranking genes based on a univariate distance statistic calculated for each gene individually. The univariate statistics summarize gene expression profiles outside of the gene co-regulation network context, leading to redundant information being included in the selection procedure. RESULTS: We propose an Eigengene-based Linear Discriminant Analysis (ELDA) to address gene selection in a multivariate framework. The algorithm uses a modified rotated Spectral Decomposition (SpD) technique to select 'hub' genes that associate with the most important eigenvectors. Using three benchmark cancer microarray datasets, we show that ELDA selects the most characteristic genes, leading to substantially smaller classifiers than the univariate feature selection based analogues. The resulting de-correlated expression profiles make the gene-wise independence assumption more realistic and applicable for the shrunken centroids classifier and other diagonal linear discriminant type of models. Our algorithm further incorporates a misclassification cost matrix, allowing differential penalization of one type of error over another. In the breast cancer data, we show false negative prognosis can be controlled via a cost-adjusted discriminant function. AVAILABILITY: R code for the ELDA algorithm is available from author upon request.

Artificial Intelligence↗

PubMed2003

Non-convulsive status epilepticus: usefulness of clinical features in selecting patients for urgent EEG.

BACKGROUND: Non-convulsive status epilepticus (NCSE) is status epilepticus without obvious tonic-clonic activity. Patients with NCSE have altered mental state. An EEG is needed to confirm the diagnosis, but obtaining an EEG on every patient with altered mental state is not practical. OBJECTIVE: To determine whether clinical features could be used to predict which patients were more likely to be in NCSE and thus in need of an urgent EEG. METHODS: Over a six month period, all patients for whom an urgent EEG was ordered to identify NCSE were enrolled. Neurology residents examined the patients and filled out a questionnaire without knowledge of the EEG results. The patients were divided into two groups, NCSE and non-NCSE, depending on the EEG result. The clinical features were compared between the two groups. The sensitivity and specificity of the features were calculated. RESULTS: 48 patients were enrolled, 12 in NCSE and 36 not in NCSE. Remote risk factors for seizures, severely impaired mental state, and ocular movement abnormalities were seen significantly more often in the NCSE group. The combined sensitivity of remote risk factors for seizures and ocular movement abnormalities was 100%. CONCLUSIONS: There are certain clinical features that are more likely to be present in patients in NCSE compared with other types of encephalopathy. Either remote risk factors for seizures or ocular movement abnormalities were seen in all patients in NCSE. These features may be used to select which patients should have an urgent EEG.

Adult↗

PubMed2006

DNA molecule classification using feature primitives.

BACKGROUND: We present a novel strategy for classification of DNA molecules using measurements from an alpha-Hemolysin channel detector. The proposed approach provides excellent classification performance for five different DNA hairpins that differ in only one base-pair. For multi-class DNA classification problems, practitioners usually adopt approaches that use decision trees consisting of binary classifiers. Finding the best tree topology requires exploring all possible tree topologies and is computationally prohibitive. We propose a computational framework based on feature primitives that eliminates the need of a decision tree of binary classifiers. In the first phase, we generate a pool of weak features from nanopore blockade current measurements by using HMM analysis, principal component analysis and various wavelet filters. In the next phase, feature selection is performed using AdaBoost. AdaBoost provides an ensemble of weak learners of various types learned from feature primitives. RESULTS AND CONCLUSION: We show that our technique, despite its inherent simplicity, provides a performance comparable to recent multi-class DNA molecule classification results. Unlike the approach presented by Winters-Hilt et al., where weaker data is dropped to obtain better classification, the proposed approach provides comparable classification accuracy without any need for rejection of weak data. A weakness of this approach, on the other hand, is the very "hands-on" tuning and feature selection that is required to obtain good generalization. Simply put, this method obtains a more informed set of features and provides better results for that reason. The strength of this approach appears to be in its ability to identify strong features, an area where further results are actively being sought.

Computational Biology↗

PubMed2005

A wrapper-based approach to image segmentation and classification.

The traditional processing flow of segmentation followed by classification in computer vision assumes that the segmentation is able to successfully extract the object of interest from the background image. It is extremely difficult to obtain a reliable segmentation without any prior knowledge about the object that is being extracted from the scene. This is further complicated by the lack of any clearly defined metrics for evaluating the quality of segmentation or for comparing segmentation algorithms. We propose a method of segmentation that addresses both of these issues, by using the object classification subsystem as an integral part of the segmentation. This will provide contextual information regarding the objects to be segmented, as well as allow us to use the probability of correct classification as a metric to determine the quality of the segmentation. We view traditional segmentation as a filter operating on the image that is independent of the classifier, much like the filter methods for feature selection. We propose a new paradigm for segmentation and classification that follows the wrapper methods of feature selection. Our method wraps the segmentation and classification together, and uses the classification accuracy as the metric to determine the best segmentation. By using shape as the classification feature, we are able to develop a segmentation algorithm that relaxes the requirement that the object of interest to be segmented must be homogeneous in some low-level image parameter, such as texture, color, or grayscale. This represents an improvement over other segmentation methods that have used classification information only to modify the segmenter parameters, since these algorithms still require an underlying homogeneity in some parameter space. Rather than considering our method as, yet, another segmentation algorithm, we propose that our wrapper method can be considered as an image segmentation framework, within which existing image segmentation algorithms may be executed. We show the performance of our proposed wrapper-based segmenter on real-world and complex images of automotive vehicle occupants for the purpose of recognizing infants on the passenger seat and disabling the vehicle airbag. This is an interesting application for testing the robustness of our approach, due to the complexity of the images, and, consequently, we believe the algorithm will be suitable for many other real-world applications.

Algorithms↗

PubMed2001

Analysis of temporal changes of mammographic features: computer-aided classification of malignant and benign breast masses.

A new classification scheme was developed to classify mammographic masses as malignant and benign by using interval change information. The masses on both the current and the prior mammograms were automatically segmented using an active contour method. From each mass, 20 run length statistics (RLS) texture features, 3 speculation features, and 12 morphological features were extracted. Additionally, 20 difference RLS features were obtained by subtracting the prior RLS features from the corresponding current RLS features. The feature space consisted of the current RLS features, the difference RLS features, the current and prior speculation features, and the current and prior mass sizes. Stepwise feature selection and linear discriminant analysis classification were used to select and merge the most useful features. A leave-one-case-out resampling scheme was used to train and test the classifier using 140 temporal image pairs (85 malignant, 55 benign) obtained from 57 biopsy-proven masses (33 malignant, 24 benign) in 56 patients. An average of 10 features were selected from the 56 training subsets: 4 difference RLS features, 4 RLS features, and 1 speculation feature from the current image, and 1 speculation feature from the prior, were most often chosen. The classifier achieved an average training Az of 0.92 and a test Az of 0.88. For comparison, a classifier was trained and tested using features extracted from the 120 current single images. This classifier achieved an average training Az of 0.90 and a test Az of 0.82. The information on the prior image significantly (p = 0.015) improved the accuracy for classification of the masses.

Algorithms↗

PubMed2001

Analysis of strain and regional variation in gene expression in mouse brain.

BACKGROUND: We performed a statistical analysis of a previously published set of gene expression microarray data from six different brain regions in two mouse strains. In the previous analysis, 24 genes showing expression differences between the strains and about 240 genes with regional differences in expression were identified. Like many gene expression studies, that analysis relied primarily on ad hoc 'fold change' and 'absent/present' criteria to select genes. To determine whether statistically motivated methods would give a more sensitive and selective analysis of gene expression patterns in the brain, we decided to use analysis of variance (ANOVA) and feature selection methods designed to select genes showing strain- or region-dependent patterns of expression. RESULTS: Our analysis revealed many additional genes that might be involved in behavioral differences between the two mouse strains and functional differences between the six brain regions. Using conservative statistical criteria, we identified at least 63 genes showing strain variation and approximately 600 genes showing regional variation. Unlike ad hoc methods, ours have the additional benefit of ranking the genes by statistical score, permitting further analysis to focus on the most significant. Comparison of our results to the previous studies and to published reports on individual genes show that we achieved high sensitivity while preserving selectivity. CONCLUSIONS: Our results indicate that molecular differences between the strains and regions studied are larger than indicated previously. We conclude that for large complex datasets, ANOVA and feature selection, alone or in combination, are more powerful than methods based on fold-change thresholds and other ad hoc selection criteria.

Analysis of Variance↗

PubMed1983

Disturbances of the blood-brain barrier in cerebrovascular disorders.

The disturbances of the BBB in cerebrovascular disorders may affect adversely an underlying basic pathological condition. Breakdown of the barrier associated with extravasation of serum proteins leads to development of vasogenic edema in the brain tissue. An abnormal passage of pharmacologically active substances, such as biogenic amines, may significantly affect cerebral blood flow and metabolism and activate neurons equipped with receptors for these substances. Also, a barrier dysfunction related to faulty out-transport of metabolites may contribute to edema and tissue damage. In cerebral ischemia, following release of arterial occlusion there can be two separate openings of the barrier: the first - occurring promptly after recirculation and related to ensuing reactive hyperemia, the second - after some delay and related to pathological changes in the brain tissue. In some circumstances, such as epileptic seizures, both "hemodynamic" and "tissue" factors may be operative at the same time. The selective features of BBB changes are related to multiplicity of barrier systems residing in cerebral endothelium. These selective features are demonstrable during development and during reversibility of postichemic barrier disturbances. Intermittent openings of the barrier observed in chronic hypertension may lead to accumulation of extravasated serum proteins and be responsible for frequently observed edematous changes in this condition.

Animals↗

PubMed2006

Cell-nuclear data reduction and prognostic model selection in bladder tumor recurrence.

OBJECTIVE: The paper aims at improving the prediction of superficial bladder recurrence. To this end, feedforward neural networks (FNNs) and a feature selection method based on unsupervised clustering, were employed. MATERIAL AND METHODS: A retrospective prognostic study of 127 patients diagnosed with superficial urinary bladder cancer was performed. Images from biopsies were digitized and cell nuclei features were extracted. To design FNN classifiers, different training methods and architectures were investigated. The unsupervised k-windows (UKW) and the fuzzy c-means clustering algorithms were applied on the feature set to identify the most informative feature subsets. RESULTS: UKW managed to reduce the dimensionality of the feature space significantly, and yielded prediction rates 87.95% and 91.41%, for non-recurrent and recurrent cases, respectively. The prediction rates achieved with the reduced feature set were marginally lower compared to the ones attained with the complete feature set. The training algorithm that exhibited the best performance in all cases was the adaptive on-line backpropagation algorithm. CONCLUSIONS: FNNs can contribute to the accurate prognosis of bladder cancer recurrence. The proposed feature selection method can remove redundant information without a significant loss in predictive accuracy, and thereby render the prognostic model less complex, more robust, and hence suitable for clinical use.

Algorithms↗

PubMed2003

Image texture analysis of sonograms in chronic inflammations of thyroid gland.

The current practice in assessing sonographic findings of chronic inflamed thyroid tissue is mainly qualitative, based just on a physician's experience. This study shows that inflamed and healthy tissues can be differentiated by automatic texture analysis of B-mode sonographic images. Feature selection is the most important part of this procedure. We employed two selection schemes for finding recognition-optimal features: one based on compactness and separability and the other based on classification error. The full feature set included Muzzolini's spatial features and Haralick's co-occurrence features. These features were selected on a set of 2430 sonograms of 81 subjects, and the classifier performance was evaluated on a test set of 540 sonograms of 18 independent subjects. A classification success rate of 100% was achieved with as few as one optimal feature among the 129 texture characteristics tested. Both selection schemes agreed on the best features. The results were confirmed on the independent test set. The stability of the results with respect to sonograph setting, thyroid gland segmentation and scanning direction was tested.

Algorithms↗

PubMed1999

Flat foot functional evaluation using pattern recognition of ground reaction data.

OBJECTIVE: Main purpose of this study was to apply quantitative gait analysis and statistical pattern recognition as clinical decision-making aids in flat foot diagnosis and post-surgery monitoring. DESIGN: Statistical pattern recognition techniques were applied to discriminate between normal and flat foot populations through ground reaction force measurements; ground reaction forces time course was assumed as a sensible index of the foot function. BACKGROUND: Gait analysis is becoming recognized as an important clinical tool in orthopaedics, in pre-surgery planning, post-surgery monitoring and in a posteriori evaluation of different treatment techniques. Statistical pattern recognition techniques have been utilized with success in this field to identify the most significant variables of selected motor functions in different pathologies, and to design classification rules and quantitative evaluation scores. METHODS: Ground reaction forces were recorded during free speed barefoot walks on 28 healthy subjects, and 28 symptomatic flexible flat foot children selected for surgical intervention. A new feature selection algorithm, based on heuristic optimization, was applied to select the most discriminant ground reaction forces time samples. A two-stage pattern recognition system, composed by three linear feature extractors, one for each ground reaction force component, and a linear classifier, was designed to classify the feet of each subject using the selected features. The output of the classifier was used to define a functional score. RESULTS: The classifier assigned the ground reaction force patterns performed by each subject into the right class with an estimated error of 15%, corresponding to an assignment error for each subject's foot of 9%. The most discriminant ground reaction forces time samples selected are in full agreement with the pathophysiology of the symptomatic flexible flat foot. The obtained score was utilized to monitor the 1 and 2 years post-operative functional recovery of two differently treated subgroups of 32 flexible flat foot subjects. CONCLUSIONS: Statistical pattern recognition techniques are promising tools for clinical gait analysis; the obtained score provides important functional information that could be used as a further aid in the clinical evaluation of flat foot and different surgical treatment techniques. RELEVANCE: Symptomatic flexible flat foot surgical decision making is frequently difficult because of the lack of objective criteria to assess functional abnormalities of the foot/ankle complex. Gait analysis and statistical pattern recognition can give us parameters with which to characterize "functional" flat foot. Moreover, we can objectively follow up the recovery of the foot/ankle complex function after surgical treatment.

Algorithms↗

PubMed2006

A QSAR model of HERG binding using a large, diverse, and internally consistent training set.

Over the past decade, the pharmaceutical industry has begun to address an addition to ADME/Tox profiling--the ability of a compound to bind to and inhibit the human ether-a-go-go-related gene (hERG)-encoded cardiac potassium channel. With the compilation of a large and diverse set of compounds measured in a single, consistent hERG channel inhibition assay, we recognized a unique opportunity to attempt to construct predictive QSAR models. Early efforts with classification models built from this training set were very encouraging. Here, we report a systematic evaluation of regression models based on neural network ensembles in conjunction with a variety of structure representations and feature selection algorithms. The combination of these modeling techniques (neural networks to capture non-linear relationships in the data, feature selection to prevent over-fitting, and aggregation to minimize model instability) was found to produce models with very good internal cross-validation statistics and good predictivity on external data.

Algorithms↗

PubMed2005

Bayesian neural network approaches to ovarian cancer identification from high-resolution mass spectrometry data.

MOTIVATION: The classification of high-dimensional data is always a challenge to statistical machine learning. We propose a novel method named shallow feature selection that assigns each feature a probability of being selected based on the structure of training data itself. Independent of particular classifiers, the high dimension of biodata can be fleetly reduced to an applicable case for consequential processing. Moreover, to improve both efficiency and performance of classification, these prior probabilities are further used to specify the distributions of top-level hyperparameters in hierarchical models of Bayesian neural network (BNN), as well as the parameters in Gaussian process models. RESULTS: Three BNN approaches were derived and then applied to identify ovarian cancer from NCI's high-resolution mass spectrometry data, which yielded an excellent performance in 1000 independent k-fold cross validations (k = 2,...,10). For instance, indices of average sensitivity and specificity of 98.56 and 98.42%, respectively, were achieved in the 2-fold cross validations. Furthermore, only one control and one cancer were misclassified in the leave-one-out cross validation. Some other popular classifiers were also tested for comparison. AVAILABILITY: The programs implemented in MatLab, R and Neal's fbm.2004-11-10.

Bayes Theorem↗

PubMed2007

Fusion algorithm for poultry skin tumor detection using hyperspectral data.

We consider a feature selection method to detect skin tumors on chicken carcasses using hyperspectral (HS) reflectance data. Detection of chicken tumors is difficult because the tumors vary in size and shape; some tumors are small, early-stage tumor spots. We make use of the fact that a chicken skin tumor consists of a lesion region surrounded by a region of thickened skin and that the spectral responses of the lesion and the thickened-skin regions of tumors are considerably different and train our feature selection algorithm to separately detect lesion regions and thickened-skin regions; we then fuse the two HS detection results to reduce false alarms. To the best of our knowledge, these techniques are new. Our forward selection and modified branch and bound algorithm is used to select a small number of lambda spectral features that are useful for discrimination. Initial results show that our method offers promise for a good tumor detection rate and a low false alarm rate.

Algorithms↗

PubMed2005

Analysis of recursive gene selection approaches from microarray data.

MOTIVATION: Finding a small subset of most predictive genes from microarray for disease prediction is a challenging problem. Support vector machines (SVMs) have been found to be successful with a recursive procedure in selecting important genes for cancer prediction. However, it is not well understood how much of the success depends on the choice of the specific classifier and how much on the recursive procedure. We answer this question by examining multiple classifers [SVM, ridge regression (RR) and Rocchio] with feature selection in recursive and non-recursive settings on three DNA microarray datasets (ALL-AML Leukemia data, Breast Cancer data and GCM data). RESULTS: We found recursive RR most effective. On the AML-ALL dataset, it achieved zero error rate on the test set using only three genes (selected from over 7000), which is more encouraging than the best published result (zero error rate using 8 genes by recursive SVM). On the Breast Cancer dataset and the two largest categories of the GCM dataset, the results achieved by recursive RR are also very encouraging. A further analysis of the experimental results shows that different classifiers penalize redundant features to different extent and this property plays an important role in the recursive feature selection process. RR classifier tends to penalize redundant features to a much larger extent than the SVM does. This may be the reason why recursive RR has a better performance in selecting genes.

Algorithms↗

PubMed2002

Discriminant histological features in the diagnosis of chronic idiopathic inflammatory bowel disease: analysis of a large dataset by a novel data visualisation technique.

BACKGROUND/AIMS: The histopathological assessment of endoscopic colorectal biopsies is important in the distinction between normality and chronic idiopathic inflammatory bowel disease, and between ulcerative colitis and Crohn's disease, in subjects with symptoms of bowel dysfunction. This study aims to use carefully defined histopathological observations on a large study population to produce systems that improve classification into these diagnostic categories. METHODS: Eight hundred and nine endoscopic colorectal biopsies with verified outcomes (165 normal, 473 ulcerative colitis, 171 Crohn's disease) were examined by a single experienced histopathologist and 20 defined features were recorded for each case using a novel graphical interface with reference images of each feature. These features, together with age and sex, were used to produce and test statistical classifiers using logistic regression and a novel growing cell structure technique. RESULTS: The distinction between chronic idiopathic inflammatory bowel disease and normality was made with a good level of performance by both statistical classifiers (with areas under the receiver operating characteristic curves above 0.80). The growing cell structure system selected features as discriminant that agreed with the published literature. Logistic regression produced a more variable selection of discriminant features because of the high correlation between many features. The distinction between ulcerative colitis and Crohn's disease was performed less accurately, with areas under the receiver operating characteristic curves of about 0.70. Again the features selected as discriminant broadly agreed with those in the published literature. CONCLUSIONS: Histopathological examination of endoscopic colorectal biopsies is an effective method of distinguishing between subjects with chronic idiopathic inflammatory bowel disease and normality, but less good at distinguishing between ulcerative colitis and Crohn's disease. The features selected as discriminant in this large statistical analysis broadly agree with those published in the literature from more qualitative studies.

Adolescent↗