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 1,153 records · Page 64Linked to original sources

PubMed2002

Predicting CNS permeability of drug molecules: comparison of neural network and support vector machine algorithms.

Two different machine-learning algorithms have been used to predict the blood-brain barrier permeability of different classes of molecules, to develop a method to predict the ability of drug compounds to penetrate the CNS. The first algorithm is based on a multilayer perceptron neural network and the second algorithm uses a support vector machine. Both algorithms are trained on an identical data set consisting of 179 CNS active molecules and 145 CNS inactive molecules. The training parameters include molecular weight, lipophilicity, hydrogen bonding, and other variables that govern the ability of a molecule to diffuse through a membrane. The results show that the support vector machine outperforms the neural network. Based on over 30 different validation sets, the SVM can predict up to 96% of the molecules correctly, averaging 81.5% over 30 test sets, which comprised of equal numbers of CNS positive and negative molecules. This is quite favorable when compared with the neural network's average performance of 75.7% with the same 30 test sets. The results of the SVM algorithm are very encouraging and suggest that a classification tool like this one will prove to be a valuable prediction approach.

Algorithms↗

PubMed1997

Application of a multi-structure neural network (MSNN) to sorting pistachio nuts.

A multi-structure neural network (MSNN) classifier consisting of four discriminators followed by a maximum selector was designed and applied to classification of four grades of pistachio nuts. Each discriminator was a multi-layer feed-forward neural network with two hidden layers and a single-neuron output layer. Fourier descriptors of the nuts' boundaries and their area were used as the recognition features. The individual discriminators were trained using a biased technique and a back-propagation algorithm. The MSNN classifier gave an average classification performance of 95.0%. This was an increase of 14.8% over the performance of a multi-layer neural network (MLNN) with similar complexity for classifying the same set of patterns.

Discrimination Learning↗

PubMed2004

A fast sequential image fractal coding approach based on optimal fuzzy clustering.

To reduce the coding time of the conventional method, a fast sequential image fractal compression algorithm was proposed on the basis of the principle of optimal fuzzy clustering (OFC) for an unsupervised sample set with the category number settled by the algorithm itself. We utilized the cost function defined by the OFC algorithm to obtain the best category number corresponding to the minimum value of the function. Firstly the Linde-Buzo-Gray (LBG) algorithm was realized to acquire a rough cluster of the domain pool. Then the optimal category number was obtained by implementing our algorithm with small computational cost. Finally the more precise category was gained and the detail of the reconstructed image efficiently preserved. As a global optimal algorithm, OFC not only helps LBG eliminate the local minima, but also effectively compensates for the arbitrary interference in hard clustering problem. Soft clustering of the domain blocks allows classified searches instead of global ones and takes less coding time, and therefore clearly outperforms to the classic method relying on reduction of the size of the domain pool by classification. In computer simulation, OFC-based algorithm for the fractal coding scheme achieved excellent performance. For some standard and sequential medical images, the results denoted that the encoding speed was improved by about 5 folds without affecting the signal-to-noise ratio and compression ratio, and the quality of the reconstructed image could be better retained.

Algorithms↗

PubMed2005

Convergence of stochastic learning in perceptrons with binary synapses.

The efficacy of a biological synapse is naturally bounded, and at some resolution, and is discrete at the latest level of single vesicles. The finite number of synaptic states dramatically reduce the storage capacity of a network when online learning is considered (i.e., the synapses are immediately modified by each pattern): the trace of old memories decays exponentially with the number of new memories (palimpsest property). Moreover, finding the discrete synaptic strengths which enable the classification of linearly separable patterns is a combinatorially hard problem known to be NP complete. In this paper we show that learning with discrete (binary) synapses is nevertheless possible with high probability if a randomly selected fraction of synapses is modified following each stimulus presentation (slow stochastic learning). As an additional constraint, the synapses are only changed if the output neuron does not give the desired response, as in the case of classical perceptron learning. We prove that for linearly separable classes of patterns the stochastic learning algorithm converges with arbitrary high probability in a finite number of presentations, provided that the number of neurons encoding the patterns is large enough. The stochastic learning algorithm is successfully applied to a standard classification problem of nonlinearly separable patterns by using multiple, stochastically independent output units, with an achieved performance which is comparable to the maximal ones reached for the task.

Algorithms↗

PubMed1998

Maximum A posteriori classification of DNA structure from sequence information.

We introduce an algorithm, LLLAMA, which combines simple pattern recognizers into a general method for estimating the entropy of a sequence. Each pattern recognizer exploits a partial match between subsequences to build a model of the sequence. Since the primary features of interest in biological sequence domains are subsequences with small variations in exact composition, LLLAMA is particularly suited to such domains. We describe two methods, LLLAMA-length and LLLAMA-alone, which use this entropy estimate to perform maximum a posteriori classification. We apply these methods to several problems in three-dimensional structure classification of short DNA sequences. The results include a surprisingly low 3.6% error rate in predicting helical conformation of oligonucleotides. We compare our results to those obtained using more traditional methods for automated generation of classifiers.

Algorithms↗

PubMed2004

Embedded image compression based on wavelet pixel classification and sorting.

The method of modeling and ordering in wavelet domain is very important to design a successful algorithm of embedded image compression. In this paper, the modeling is limited to "pixel classification," the relationship between wavelet pixels in significance coding. Similarly, the ordering is limited to "pixel sorting," the coding order of wavelet pixels. We use pixel classification and sorting to provide a better understanding of previous works. The image pixels in wavelet domain are classified and sorted, either explicitly or implicitly, for embedded image compression. A new embedded image code is proposed based on a novel pixel classification and sorting (PCAS) scheme in wavelet domain. In PCAS, pixels to be coded are classified into several quantized contexts based on a large context template and sorted based on their estimated significance probabilities. The purpose of pixel classification is to exploit the intraband correlation in wavelet domain. Pixel sorting employs several fractional bit-plane coding passes to improve the rate-distortion performance. The proposed pixel classification and sorting technique is simple, yet effective, producing an embedded image code with excellent compression performance. In addition, our algorithm is able to provide either spatial or quality scalability with flexible complexity.

Algorithms↗

PubMed2002

Development of an HCV infection risk stratification algorithm for patients on chronic hemodialysis.

OBJECTIVE: The prevalence of hepatitis C virus (HCV) in patients on chronic hemodialysis (HD) is near 9%. Transaminases, which are lower in HD patients, are not effective in screening for HCV. Our aim was to design an HCV risk stratification strategy incorporating lowered aminotransferase levels and other clinical parameters. METHODS: Patient serum from 168 consecutive HD patients was analyzed for AST, ALT, ferritin, and hepatitis C antibody. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated for lower transaminase values. Multivariate classification and regression tree analysis was used to determine the best combination of variables to predict risk for HCV infection. RESULTS: Median AST and ALT levels were higher in anti-HCV Ab(+) patients (p < 0.05). Applying a lower cutoff value for ALT of 16 IU/L resulted in a sensitivity of 61.1%, a specificity of 66.7%, a positive predictive value of 33.9%, and a negative predictive value of 86.0% for detection of HCV infection. Multivariate classification and regression tree analysis derived an algorithm using patient age, months on HD, and AST, resulting in a 97.2% sensitivity and a 51.9% specificity for the detection of HCV(+) HD patients. CONCLUSIONS: A lower normal cutoff value of 18 IU/L for AST and 16 IU/L for ALT increased sensitivity and specificity for the detection of HCV infection in HD patients. An algorithm combining lower transaminases with clinical parameters improved both sensitivity and specificity in HCV detection. Prospective confirmation of this algorithm would allow more selective HCV enzyme immunoassay and polymerase chain reaction testing in dialysis units.

Alanine Transaminase↗

PubMed1985

The research diagnostic criteria for endogenous depression and the dexamethasone suppression test: a discriminant function analysis.

Most studies examining the validity of the Research Diagnostic Criteria (RDC) for endogenous depression have been negative. RDC endogenous subtyping is not associated with short- or long-term treatment outcome, family history of affective disorder, or premorbid personality disorder. Studies examining its relationship to the dexamethasone suppression test (DST) are mixed; half report a significant association, and half do not. The RDC endogenous diagnosis may lack validity either because the criteria do not represent, or are not specific to, the endogenous subtype, or the diagnostic algorithm is inappropriate. In the present study, we conduct a discriminant function analysis on the 10 criteria for the endogenous subtype using DST results as the independent variable. We constructed a new diagnostic algorithm and cross-validated it on a second patient sample. In both samples the discriminant function classification was significantly associated with DST results, whereas the RDC algorithm was not.

Adult↗

PubMed2005

Spondylolysis and spondylolisthesis in the child and adolescent: a new classification.

Spondylolysis and spondylolisthesis commonly are diagnosed in children and adolescents. The diagnostic workup and treatment plan vary widely among physicians. Although the orthopaedic literature is extensive on the topic, it is our opinion that a lack of clarity exists with regards to etiology, terminology, subtypes of spondylolysis and spondylolisthesis, and treatment. Important basic principles regarding spondylolysis and spondylolisthesis, with emphasis on clinical evaluation and nonsurgical treatment, serve as the basis for a new classification. We propose a new classification for pediatric spondylolysis and spondylolisthesis that is comprehensive, simple, and easily applied. This scheme is based on clinical presentation and spinal morphology and is more appropriate for the child and adolescent than the existing classification schemes of Wiltse-Newman and Marchetti-Bartolozzi. Algorithms for evaluation and treatment of spondylolysis and spondylolisthesis in children and adolescents, based on this new classification, are presented.

Adolescent↗

PubMed1998

Self-organizing tree-growing network for the classification of protein sequences.

The self-organizing tree algorithm (SOTA) was recently introduced to construct phylogenetic trees from biological sequences, based on the principles of Kohonen's self-organizing maps and on Fritzke's growing cell structures. SOTA is designed in such a way that the generation of new nodes can be stopped when the sequences assigned to a node are already above a certain similarity threshold. In this way a phylogenetic tree resolved at a high taxonomic level can be obtained. This capability is especially useful to classify sets of diversified sequences. SOTA was originally designed to analyze pre-aligned sequences. It is now adapted to be able to analyze patterns associated to the frequency of residues along a sequence, such as protein dipeptide composition and other n-gram compositions. In this work we show that the algorithm applied to these data is able to not only successfully construct phylogenetic trees of protein families, such as cytochrome c, triosephophate isomerase, and hemoglobin alpha chains, but also classify very diversified sequence data sets, such as a mixture of interleukins and their receptors.

Algorithms↗

PubMed2004

[Discussion of naive Bayesian algorithm in prognosis prediction of primary liver cancer].

OBJECTIVE: To apply naive Bayesian algorithm in prognosis prediction of primary liver cancer and to predict the survival expectation of patients after transcatheter arterial chemoembolization (TACE). METHOD: Naive Bayesian algorithm was applied. Using correlation analysis to sift data-attributes. Whereas the missing data were assumed to follow the same distribution as that of the known. RESULT: The same-distribution assumption of the missing data reduces the error rate from 71.9% to 9.4%. Twelve attributes were sifted from 39 attributes by the correlation analysis, which were more effective to the final classification, and had a relatively low error rate of 3.1%. CONCLUSION: The proposed method effectively increases the accuracy of classification. Successful application of the naive Bayesian algorithm in prognostic problem of primary liver cancer indicates a bright future of this method in medical field.

Algorithms↗

PubMed1986

Applications of computerized interactive morphometry in pathology. II. A model for computer generated diagnosis.

We present a model for the analysis and tentative diagnosis of pathologic problems by a trained observer utilizing data generated by a video based computerized interactive morphometry system in conjunction with multivariate methods of discriminant classificatory analysis that separate classes based on unweighted numerical values and with adhoc algorithms based on hierarchical analysis. The model was tested with two diagnostic problems that could benefit from a morphometric approach: the classification of non-Hodgkin's lymphomas from routine histologic slides and the distinction of malignant mesotheliomas from benign effusions in smears prepared from pleural effusions. The touch-sensitive screen of the computerized interactive morphometry system enables trained observers to measure the real time image of profiles of interest either by using graphic standards or simply by touching the two extreme points of the diameter of interest. This procedure generates multiple data in the form of classes that can be effectively used toward the more objective discrimination of lesions of difficult classification by analysis with a combination of adhoc diagnostic algorithms and multivariate statistical methods.

Computers↗

PubMed1999

Artifact processing during exercise testing.

In signal processing of exercise electrocardiograms (ECGs), artifacts are a recurring problem. It is still difficult to discriminate the ECG curves from artifacts, especially in exercise ECGs and particularly in the high exercise phase. We focused on the artifact problem and worked on two new topics: the Finite Impulse Response Residual Filtering (FRF) algorithm and the Intelligent Lead Switch algorithm. The FRF algorithm reduces the baseline wander and muscle noise in the ECG stream, with much less distortion of the QRS complexes. It subtracts a continuously updated median beat from the current ECG, filters the residual signal with a high-pass and a low-pass filter, and adds the median beat to the filtered residual signal. The Intelligent Lead Switch algorithm takes advantage of the redundancy of a multilead system (eg, standard leads), which is nowadays used during exercise testing. It selects the best leads for QRS detection and thus improves the heart rate calculation, ST segment evaluation, and arrhythmia classification.

Algorithms↗

PubMed2000

Efficient perceptron learning using constrained steepest descent.

An algorithm is proposed for training the single-layered perceptron. The algorithm follows successive steepest descent directions with respect to the perceptron cost function, taking care not to increase the number of misclassified patterns. The problem of finding these directions is stated as a quadratic programming task, to which a fast and effective solution is proposed. The resulting algorithm has no free parameters and therefore no heuristics are involved in its application. It is proved that the algorithm always converges in a finite number of steps. For linearly separable problems, it always finds a hyperplane that completely separates patterns belonging to different categories. Termination of the algorithm without separating all given patterns means that the presented set of patterns is indeed linearly inseparable. Thus the algorithm provides a natural criterion for linear separability. Compared to other state of the art algorithms, the proposed method exhibits substantially improved speed, as demonstrated in a number of demanding benchmark classification tasks.

Algorithms↗

PubMed1998

The subclass approach for mutational spectrum analysis: application of the SEM algorithm.

Analysis and comparison of mutational spectra represents an important problem in molecular biology. To analyse a mutational spectra we apply an algorithm based on the SEM subclass approach (Simulation, Expectation, Maximization). The algorithm tries to classify the mutational sites according to different mutation probabilities, and each site should belong to one class. Each class is approximated by binomial distribution and thus any real mutational spectrum is regarded as a mixture of binomial distributions. The separation process runs iteratively. Each iteration includes the simulation, maximization and estimation procedures. To evaluate the quality of the classification results, the X2 test is used. The algorithm has been checked on random spectra with preset parameters and on real mutational spectra. As has been shown, 17 out of 19 analysed real mutational spectra can be divided into two or more classes of sites, of which one contains hotspots of mutation. For the G:C-->A:T mutational spectra induced by Sn1 alkylating mutagenes (11 spectra) the classification accuracy was 0.95. To test different site volumes, each Sn1-induced spectrum was divided into the G-->A and C-->T spectra. The classification accuracy for these spectra was 0.96. From the analysis of classification errors it is possible to suggest that at least part of them cannot be ascribed to the faults of the algorithm but are caused by some special features of the mutagenesis itself. The results of the real data are in good relation with existing knowledge. The approach we present is an attempt to formalize the concept of a "mutational hotspot". The program implementing the SEM algorithm is available on the Web server (http:/(/)www.itba.mi.cnr.it/webmutation).

Algorithms↗

PubMed2003

Noninvasive study of ventricular preexcitation using multichannel magnetocardiography.

In clinical practice, noninvasive classification of ventricular preexcitation (VPX) is usually done with ECG algorithms, which provide only a qualitative localization of accessory pathways. Since 1984, single or multichannel magnetocardiography (MMCG) has been used for three-dimensional localization of VPX sites, but a systematic study comparing the results of ECG and MMCG methods was lacking. This study evaluated the reliability of MMCG in an unshielded electrophysiological catheterization laboratory, and compared VPX classification as achieved with the five most recent ECG algorithms with that obtained by MMCG mapping and imaging techniques. A nine-channel direct current superconducting quantum interference device (DC-SQUID) MMCG system (sensitivity is 20 fT/Hz0.5) was used for sequential MMCG from 36 points on the anterior chest wall, within an area 20 x 20 cm. Twenty-eight patients with Wolff-Parkinson-White syndrome were examined at least twice, on the same day or after several months to test the reproducibility of the measurements. In eight patients, the reproducibility of MMCG was also evaluated using different MCG instrumentation during maximal VPX and/or atrioventricular reentrant tachycardia induced by transesophageal atrial pacing via a nonmagnetic catheter. The results of VPX localization with ECG algorithms and MMCG were compared. Equivalent current dipole, effective magnetic dipole, and distributed currents imaging models were used for the inverse solution. MMCG classification of VPX was found to be more accurate than ECG methods, and also provided additional information for the identification of paraseptal pathways. Furthermore, in patients with complex activation patterns during the delta wave, distributed currents imaging revealed two different activation patterns, suggesting the existence of multiple accessory pathways.

Algorithms↗

PubMed2001

[Artificial intelligence in sleep analysis (ARTISANA)--modelling visual processes in sleep classification].

We describe a novel approach to the problem of automated sleep stage recognition. The ARTISANA algorithm mimics the behaviour of a human expert visually scoring sleep stages (Rechtschaffen and Kales classification). It comprises a number of interacting components that imitate the stepwise approach of the human expert, and artificial intelligence components. On the basis of parameters extracted at 1-s intervals from the signal curves, artificial neural networks recognize the incidence of typical patterns, e.g. delta activity or K complexes. This is followed by a rule interpretation stage that identifies the sleep stage with the aid of a neuro-fuzzy system while taking account of the context. Validation studies based on the records of 8 patients with obstructive sleep apnoea have confirmed the potential of this approach. Further features of the system include the transparency of the decision-taking process, and the flexibility of the option for expanding the system to cover new patterns and criteria.

Adult↗

PubMed2005

Automated classification of an environmental sensitivity index.

Environmental Sensitivity Indices (ESI) composed of many field-data are essential for monitoring and control systems. At the beginning of the last decade an ESI of the German Wadden Sea was developed for use by the relevant authorities. This ESI was derived by experts semi-manually analysing the extensive field data-set. An algorithm is presented here which emulates human expert-decisions on the classification of sensitivity classes. This will permit the necessary regular updates of ESI-determination when new field data become available using automated classifications procedures. After tuning the algorithm parameters it generates decisions identical to those of human experts in about 97% of all locations tested. In addition, the algorithm presented also enables erroneous or extremely seldom field data to be identified.

Algorithms↗