PubMed Health⌕ Search

SEARCH · PubMed Health

Results for “Classification Algorithms”

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 883 records · Page 49Linked to original sources
PubMed1996

[A new classification of the motor disorders in patients who have had a cerebral stroke].

A new classification of motor disorders in patients after brain hemisphere's stroke as well as with its sequelae was proposed on the basis of clinical electromyographic studies. The classification validity was confirmed by mathematic statistic methods. The classification was composed of motor syndromes and is based on the following criteria: the severity of paresis, the correlation between gravity of paresis of the upper and lower limbs, muscular spasticity, alteration of integral estimation of active movements (the motor algorithm). The most informative signs of classification are the following: the paresis gravity, the degree of muscular spasticity, the alteration of motor algorithm, the parameters of stimulative electromyography. The described classification permits to carry out differentiated actions of neurorehabilitation directed to intensification of effectivity of treatment after hemisphere's stroke.

Algorithms↗
PubMed2006

A clinically applicable fracture classification for distal humeral fractures.

The purpose of this study was to design a clinically applicable classification for distal humeral fractures that would provide guidance to the surgeon with regard to surgical approach and operative management. The new classification was assessed by use of the original radiographs from a study comparing distal humeral fracture classifications undertaken in Oxford, England, and was validated by use of the exact methodology of that study. Nine independent assessors were asked to classify 33 sets of radiographs on 2 separate occasions using the classifications of Riseborough and Radin, Mehne and Jupiter, and the AO, as well as the new classification system. With the use of the kappa statistic, the level of interobserver and intraobserver agreement was determined. The new classification system was found to be both substantially reliable (kappa, 0.664) and reproducible (kappa, 0.732). The new classification achieved superior interobserver and intraobserver agreement compared with the other 3 classification systems, with a low proportion of unclassifiable fractures. Used in conjunction with a management algorithm, we believe that the new classification aids the surgical decision-making process for these complex fractures.

Adult↗
PubMed2003

Microarray-based cancer diagnosis with artificial neural networks.

In recent years, the advent of experimental methods to probe gene expression profiles of cancer on a genome-wide scale has led to widespread use of supervised machine learning algorithms to characterize these profiles. The main applications of these analysis methods range from assigning functional classes of previously uncharacterized genes to classification and prediction of different cancer tissues. This article surveys the application of machine learning algorithms to classification and diagnosis of cancer based on expression profiles. To exemplify the important issues of the classification procedure, the emphasis of this article is on one such method, namely artificial neural networks. In addition, methods to extract genes that are important for the performance of a classifier, as well as the influence of sample selection on prediction results are discussed.

Algorithms↗
PubMed2004

Determination of menopausal status in women: the NHLBI-sponsored Women's Ischemia Syndrome Evaluation (WISE) Study.

PURPOSE: Accurate classification of menopausal status is important to epidemiological research evaluating the role of reproductive hormones in disease processes. Algorithms relying on repeat hormone assays are unfeasible in large epidemiological studies. This paper summarizes the development of the Women's Ischemia Syndrome Evaluation (WISE) Hormonal menopausal status algorithm for determining premenopausal, perimenopausal, and postmenopausal status using menstrual and reproductive history and reproductive hormone levels obtained at a single clinic visit. METHODS: The authors compared the accuracy of this algorithm with two currently used self-report algorithms: Menstrual, based only on months since last menstrual period, and Historical, which adds age and surgical history. RESULTS: The study population consisted of 515 women (329 clearly postmenopausal) enrolled in the WISE study who were undergoing coronary angiography for suspected ischemia. A subgroup of 186, not clearly postmenopausal, was classified by these three algorithms. Results were evaluated against individualized expert consensus classification. The Menstrual and Historical classifications differed significantly (p < 0.0001) from expert consensus, with 32%-26% discordant classifications, respectively. For the WISE Hormonal classification, discordance was 4%. CONCLUSIONS: The authors conclude that inaccurate classifications of menopausal status occur frequently in self-report algorithms. Use of the relatively simple WISE algorithm can improve the accuracy of menopausal status classification for epidemiological research.

Adult↗
PubMed1985

Cluster analysis to improve food classification within commodity groups.

Mathematical clustering algorithms were used to classify foods within dairy, grain, and fat commodity groups on the basis of nutrients with limited availability in the food supply as well as those posing a possible health risk due to excess consumption. The procedure overcomes the problem that has made objective and accurate grouping, i.e., dealing simultaneously with 10 or more nutrients, difficult. The clustering routine classifies foods on the basis of similar nutrient content for any number of food attributes and assigns a degree of association to each food to indicate its compositional similarity to a prototype food for the cluster group. Foods within dairy, grain, and fat commodity groups were clustered on the basis of similar content of vitamin B-6, calcium, iron, magnesium, folacin, zinc, and added sugar, fat, cholesterol, and sodium. Whole milk and natural cheese clustered together on the basis of their moderate nutrient and relatively high fat and sodium content. Whole wheat breads, pumpernickel bread, and pancakes from mix constituted a grain subgroup with highest nutrient content, lowest cholesterol and sugar, lower fat, and higher sodium. Other subgroups based upon similarities in attributes were identified within food commodity categories. The result is an expansion of some food groups to incorporate concepts of both nutritional adequacy and moderation of food components of current nutritional concern.

Dairy Products↗
PubMed2006

Augur--a computational pipeline for whole genome microbial surface protein prediction and classification.

UNLABELLED: The analysis of protein function is a challenge and a major bottleneck towards well-annotated and analysed microbial genomes. In particular, bacterial surface proteins present an opportunity for pharmacological intervention and vaccine development. We present Augur, an automatic prediction pipeline that integrates major surface prediction algorithms and enables comparative analysis, classification and visualization for gram-positive bacteria on a genomic scale. AVAILABILITY: http://bioinfo.mikrobio.med.uni-giessen.de/augur

Algorithms↗
PubMedDate not supplied

Integrated homology modelling and X-ray study of herpes simplex virus I thymidine kinase: a case study.

Knowledge-based homology modelling together with site-directed mutagenesis, epitope and conformational mapping is an approach to predict the structures of proteins and for the rational design of new drugs. In this study we present how this procedure has been applied to model the structure of herpes simplex virus type 1 thymidine kinase (HSV1 TK, HSV1 ATP-thymidine-5'-phosphotransferase, EC 2.7.1.21). We have used, and evaluated, several secondary structure prediction methods, such as the classical one based on Chou and Fastman algorithm, neural networks using the Kabsch and Sander classification, and the PRISM method. We have validated the algorithms by applying them to the porcine adenylate kinase (ADK), whose three-dimensional structure is known and that has been used for the alignment of the TKs as well. The resulting first model of HSV1-TK consisted of the first beta-strand connected to the phosphate binding loop and its subsequent alpha-helix, the fourth beta-strand connected to the conserved FDRH sequence and two alpha-helix with basic amino acids. The 3D structure was built using the X-ray structure of ADK as template and following the general procedure for homology modelling. We extended the model by means of COMPOSER, an automatic process for protein modelling. Site-directed mutagenesis was used to experimentally verify the predicted active-site model of HSV1-TK. The data measured in our lab and by others support the suggestion that the FDRH motif is part of the active site and plays an important role in the phosphorylation of substrates. The structure of HSV1 TK, recently solved in collaboration with Prof. G. Schulz at 2.7 A resolution, includes 284 of 343 residues of the N-terminal truncated TK. The secondary structures could be clearly assigned and fitted to the density. The comparison between crystallographically determined structure and the model shows that nearly 70% of the HSV1 TK structure has been correctly modelled by the described integrated approach to knowledge based ligand protein complex structure prediction. This indicate that computer assisted methods, combined with "manual" correction both for alignment and 3D construction are useful and can be successful.

Crystallography, X-Ray↗
PubMed1989

Diagnostic criteria for hospitalized acute myocardial infarction: the Minnesota experience.

Standardized diagnostic algorithms are needed for systematic surveillance of hospitalized acute myocardial infarction (AMI). Ambiguities in diagnostic classification are resolvable to the extent that objective information is available in the hospital chart. In this study of diagnostic algorithms, serum cardiac enzyme levels, especially creatine kinase total (CK-TOT) and creatine kinase myocardial band (CK-MB) isoenzyme, were most closely correlated with the physician-reviewer diagnostic assignment used for validation; chest pain and electrocardiographic findings were less closely correlated. In addition, a close relationship was noted between the clinician's diagnostic impression and testing procedures and the final hospital discharge diagnosis. Thus, the algorithm should include discharge diagnosis as a classification element. The algorithm for cases discharged as acute myocardial infarction should be very sensitive, tending to call cases acute myocardial infarction. Other discharge diagnoses may harbour some clinically unrecognized myocardial infarction cases; however, the algorithm for such cases should be restrictive and specific to minimize false positives. These findings indicate optimal ways of combining clinical characteristics to most completely and accurately identify cases of acute myocardial infarction based on hospital records examined in retrospect.

Algorithms↗
PubMed2006

Imputation methods to improve inference in SNP association studies.

Missing single nucleotide polymorphisms (SNPs) are quite common in genetic association studies. Subjects with missing SNPs are often discarded in analyses, which may seriously undermine the inference of SNP-disease association. In this article, we develop two haplotype-based imputation approaches and one tree-based imputation approach for association studies. The emphasis is to evaluate the impact of imputation on parameter estimation, compared to the standard practice of ignoring missing data. Haplotype-based approaches build on haplotype reconstruction by the expectation-maximization (EM) algorithm or a weighted EM (WEM) algorithm, depending on whether case-control status is taken into account. The tree-based approach uses a Gibbs sampler to iteratively sample from a full conditional distribution, which is obtained from the classification and regression tree (CART) algorithm. We employ a standard multiple imputation procedure to account for the uncertainty of imputation. We apply the methods to simulated data as well as a case-control study on developmental dyslexia. Our results suggest that imputation generally improves efficiency over the standard practice of ignoring missing data. The tree-based approach performs comparably well as haplotype-based approaches, but the former has a computational advantage. The WEM approach yields the smallest bias at a price of increased variance.

Algorithms↗
PubMedDate not supplied

Metatarsophalangeal joint capsule tears: an analysis by arthrography, a new classification system and surgical management.

Metatarsalgia is a common presenting symptom with an established list of differential diagnoses. The authors present a classification system and surgical treatment algorithm for chronic metatarsophalangeal pain due to metatarsophalangeal joint capsule tear. A series of 58 metatarsophalangeal joints with partial tear diagnosed by arthrogram and treated by surgical repair are reviewed. The authors propose a classification system based on preoperative arthrography and a surgical repair procedure for each type of three distinct patterns. A study was developed and funded to perform postoperative arthrograms on 15 patients who had undergone surgical repair using the procedures presented. The purpose of the study was to validate the utility of the arthrogram in the diagnosis and clarification of the nature of the capsular tear. The authors were also able to demonstrate that the arthrographic findings became normal postoperatively, and that surgical repair of a seemingly innocuous capsule tear relieves pain. Fifty-six patients in the series reported relief of their preoperative symptoms. Postoperative arthrograms in 15 patients demonstrated a normal pattern in 73%, 20% had decreased extravasation, and 7% were unchanged.

Adult↗
PubMed2005

Toward automated segmentation of the pathological lung in CT.

Conventional methods of lung segmentation rely on a large gray value contrast between lung fields and surrounding tissues. These methods fail on scans with lungs that contain dense pathologies, and such scans occur frequently in clinical practice. We propose a segmentation-by-registration scheme in which a scan with normal lungs is elastically registered to a scan containing pathology. When the resulting transformation is applied to a mask of the normal lungs, a segmentation is found for the pathological lungs. As a mask of the normal lungs, a probabilistic segmentation built up out of the segmentations of 15 registered normal scans is used. To refine the segmentation, voxel classification is applied to a certain volume around the borders of the transformed probabilistic mask. Performance of this scheme is compared to that of three other algorithms: a conventional, a user-interactive and a voxel classification method. The algorithms are tested on 10 three-dimensional thin-slice computed tomography volumes containing high-density pathology. The resulting segmentations are evaluated by comparing them to manual segmentations in terms of volumetric overlap and border positioning measures. The conventional and user-interactive methods that start off with thresholding techniques fail to segment the pathologies and are outperformed by both voxel classification and the refined segmentation-by-registration. The refined registration scheme enjoys the additional benefit that it does not require pathological (hand-segmented) training data.

Algorithms↗
PubMed2006

Hierarchical stochastic image grammars for classification and segmentation.

We develop a new class of hierarchical stochastic image models called spatial random trees (SRTs) which admit polynomial-complexity exact inference algorithms. Our framework of multitree dictionaries is the starting point for this construction. SRTs are stochastic hidden tree models whose leaves are associated with image data. The states at the tree nodes are random variables, and, in addition, the structure of the tree is random and is generated by a probabilistic grammar. We describe an efficient recursive algorithm for obtaining the maximum a posteriori estimate of both the tree structure and the tree states given an image. We also develop an efficient procedure for performing one iteration of the expectation-maximization algorithm and use it to estimate the model parameters from a set of training images. We address other inference problems arising in applications such as maximization of posterior marginals and hypothesis testing. Our models and algorithms are illustrated through several image classification and segmentation experiments, ranging from the segmentation of synthetic images to the classification of natural photographs and the segmentation of scanned documents. In each case, we show that our method substantially improves accuracy over a variety of existing methods.

Algorithms↗
PubMed1998

Bayesian protein family classifier.

A Bayesian procedure for the simultaneous alignment and classification of sequences into subclasses is described. This Gibbs sampling algorithm iterates between an alignment step and a classification step. It employs Bayesian inference for the identification of the number of conserved columns, the number of motifs in each class, their size, and the size of the classes. Using Bayesian prediction, inter-class differences in all these variables are brought to bare on the classification. Application to a superfamily of cyclic nucleotide-binding proteins identifies both similarities and differences in the sequence characteristics of the five subclasses identified by the procedure: 1) cNMP-dependent kinases, 2) prokaryotic cAMP-dependent regulatory proteins, CRP-type, 3) prokaryotic regulatory proteins, FNR-type, 4) cAMP gated ion channel proteins of animals, and 5) cAMP gated ion channels of plants.

Algorithms↗
PubMed1997

Supervised Learning Extensions to the CLAM Network.

The contextual layered associative memory (CLAM) has been developed as a self-generating structure which implements a probabilistic encoding scheme. The training algorithms are geared towards the unsupervised generation of a layerable associative mapping ([Thacker and Mayhew, 1989]). We show here that the resulting structure will support layers which can be trained to produce outputs that approximate conditional probabilities of classification. Unsupervised and supervised learning algorithms operate independently permitting the unsupervised representational layer to be developed before supervision is available. The system thus supports learning which is inherently more flexible than conventional node labelling schemes. Copyright 1997 Elsevier Science Ltd. All Rights Reserved.

Journal Article↗
PubMed2001

[Radiologic screening for lung cancer: present status and future perspectives].

Radiologic screening for lung cancer: present status and future perspectives. Lung cancer is the most common cause of death from malignancy. This is predominantly due to the poor prognosis of the mostly advanced tumor stages at the time of presentation. Prognosis of early, usually asymptomatic stages is more favourable, particularly in non-small-cell histologic types. Therefore, early detection using diagnostic tests promises reduction of mortality from lung cancer. Due to its high sensitivity for small pulmonary nodules - the most common manifestation of early lung cancer - computed tomography appears suitable as a screening test particularly as the high radiation exposure associated with standard examination protocols can be significantly reduced for this purpose. Due to the high prevalence of benign small pulmonary nodules diagnostic algorithms are required for non-invasive classification of detected nodules. Preliminary studies of low-dose CT using algorithms based on size and density of detected nodules revealed a high proportion of asymptomatic lung cancers and early resectable tumor stages with a small number of invasive procedures for benign nodules. Prior to a wide application of this technique in clinical routine more data is required as to appropriate inclusion criteria, screening intervals and most importantly the effect of screening on reduction of mortality from lung cancer.

Adult↗
PubMed1997

Community-acquired pneumonia: can it be defined with claims data?

The use of administrative data to study pneumonia is limited because International Classification of Diseases, 9th revision, Clinical Modification (ICD9-CM) diagnosis codes do not specify whether pneumonia is community-acquired (CAP), a key clinical distinction. We classified 212 patients discharged with a diagnosis code for pneumonia as to whether or not they had CAP, using three administrative data-based systems (Diagnosis Related Groups (DRGs) alone, principal diagnosis alone, and a complex algorithm). We examined agreement with classification by clinician chart review. We also compared the length of stay (LOS) and mortality among the CAP populations identified with different methods. Agreement between the clinical review and the three administrative data methods ranged from 86 to 80%. Classification by DRG performed least well. Populations defined by claims data had similar mortality but shorter mean LOS (9.70, 9.40, and 7.91 days for the algorithm, principal diagnosis and DRG methods, respectively) than the clinically defined population (10.85 days). We conclude that studies of CAP using populations identified by claims may underestimate LOS.

Adult↗
PubMed2002

Strong feature sets from small samples.

For small samples, classifier design algorithms typically suffer from overfitting. Given a set of features, a classifier must be designed and its error estimated. For small samples, an error estimator may be unbiased but, owing to a large variance, often give very optimistic estimates. This paper proposes mitigating the small-sample problem by designing classifiers from a probability distribution resulting from spreading the mass of the sample points to make classification more difficult, while maintaining sample geometry. The algorithm is parameterized by the variance of the spreading distribution. By increasing the spread, the algorithm finds gene sets whose classification accuracy remains strong relative to greater spreading of the sample. The error gives a measure of the strength of the feature set as a function of the spread. The algorithm yields feature sets that can distinguish the two classes, not only for the sample data, but for distributions spread beyond the sample data. For linear classifiers, the topic of the present paper, the classifiers are derived analytically from the model, thereby providing an enormous savings in computation time. The algorithm is applied to cancer classification via cDNA microarrays. In particular, the genes BRCA1 and BRCA2 are associated with a hereditary disposition to breast cancer, and the algorithm is used to find gene sets whose expressions can be used to classify BRCA1 and BRCA2 tumors.

Breast Neoplasms↗