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 1,045 records · Page 58Linked to original sources

PubMed2004

Discriminative learning quadratic discriminant function for handwriting recognition.

In character string recognition integrating segmentation and classification, high classification accuracy and resistance to noncharacters are desired to the underlying classifier. In a previous evaluation study, the modified quadratic discriminant function (MQDF) proposed by Kimura et al. was shown to be superior in noncharacter resistance but inferior in classification accuracy to neural networks. This paper proposes a discriminative learning algorithm to optimize the parameters of MQDF with aim to improve the classification accuracy while preserving the superior noncharacter resistance. We refer to the resulting classifier as discriminative learning QDF (DLQDF). The parameters of DLQDF adhere to the structure of MQDF under the Gaussian density assumption and are optimized under the minimum classification error (MCE) criterion. The promise of DLQDF is justified in handwritten digit recognition and numeral string recognition, where the performance of DLQDF is comparable to or superior to that of neural classifiers. The results are also competitive to the best ones reported in the literature.

Discrimination Learning↗

PubMed2005

Gene selection based on multi-class support vector machines and genetic algorithms.

Microarrays are a new technology that allows biologists to better understand the interactions between diverse pathologic state at the gene level. However, the amount of data generated by these tools becomes problematic, even though data are supposed to be automatically analyzed (e.g., for diagnostic purposes). The issue becomes more complex when the expression data involve multiple states. We present a novel approach to the gene selection problem in multi-class gene expression-based cancer classification, which combines support vector machines and genetic algorithms. This new method is able to select small subsets and still improve the classification accuracy.

Algorithms↗

PubMed2006

Algorithmic computation of knot polynomials of secondary structure elements of proteins.

The classification of protein structures is an important and still outstanding problem. The purpose of this paper is threefold. First, we utilize a relation between the Tutte and homfly polynomial to show that the Alexander-Conway polynomial can be algorithmically computed for a given planar graph. Second, as special cases of planar graphs, we use polymer graphs of protein structures. More precisely, we use three building blocks of the three-dimensional protein structure--alpha-helix, antiparallel beta-sheet, and parallel beta-sheet--and calculate, for their corresponding polymer graphs, the Tutte polynomials analytically by providing recurrence equations for all three secondary structure elements. Third, we present numerical results comparing the results from our analytical calculations with the numerical results of our algorithm-not only to test consistency, but also to demonstrate that all assigned polynomials are unique labels of the secondary structure elements. This paves the way for an automatic classification of protein structures.

Algorithms↗

PubMed2004

Using amino acid patterns to accurately predict translation initiation sites.

The translation initiation site (TIS) prediction problem is about how to correctly identify TIS in mRNA, cDNA, or other types of genomic sequences. High prediction accuracy can be helpful in a better understanding of protein coding from nucleotide sequences. This is an important step in genomic analysis to determine protein coding from nucleotide sequences. In this paper, we present an in silico method to predict translation initiation sites in vertebrate cDNA or mRNA sequences. This method consists of three sequential steps as follows. In the first step, candidate features are generated using k-gram amino acid patterns. In the second step, a small number of top-ranked features are selected by an entropy-based algorithm. In the third step, a classification model is built to recognize true TISs by applying support vector machines or ensembles of decision trees to the selected features. We have tested our method on several independent data sets, including two public ones and our own extracted sequences. The experimental results achieved are better than those reported previously using the same data sets. Our high accuracy not only demonstrates the feasibility of our method, but also indicates that there might be "amino acid" patterns around TIS in cDNA and mRNA sequences.

Algorithms↗

PubMed2006

Large-scale data exploration with the hierarchically growing hyperbolic SOM.

We introduce the Hierarchically Growing Hyperbolic Self-Organizing Map (H2SOM) featuring two extensions of the HSOM (hyperbolic SOM): (i) a hierarchically growing variant that allows for incremental training with an automated adaptation of lattice size to achieve a prescribed quantization error and (ii) an approximate best match search that utilizes the special structure of the hyperbolic lattice to achieve a tremendous speed-up for large map sizes. Using the MNIST and the Reuters-21578 database as benchmark datasets, we show that the H2SOM yields a highly efficient visualization algorithm that combines the virtues of the SOM with extremely rapid training and low quantization and classification errors.

Algorithms↗

PubMedDate not supplied

A new method for lumbar herniated inter-vertebral disc diagnosis based on image analysis of transverse sections.

This paper describes an image analysis method that uses automatic algorithms for the evaluation of herniation classification and geometry in the diagnosis of lumbar herniated inter-vertebral disc (HIVD). The method uses boundary approximation that uses a B-spline curve to approximate circle-like disc boundary and excludes the herniation from other normal parts of the disc boundary and, feature recognition that classifies the herniation, and herniation shape reconstruction that infers the 3D geometry from one or more transverse sections. This method can be used as a qualitative and quantitative tool for the diagnosis of lumbar HIVD using transverse sections.

Adult↗

PubMed2002

Automated breath detection on long-duration signals using feedforward backpropagation artificial neural networks.

A new breath-detection algorithm is presented, intended to automate the analysis of respiratory data acquired during sleep. The algorithm is based on two independent artificial neural networks (ANN(insp) and ANN(expi)) that recognize, in the original signal, windows of interest where the onset of inspiration and expiration occurs. Postprocessing consists in finding inside each of these windows of interest minimum and maximum corresponding to each inspiration and expiration. The ANN(insp) and ANN(expi) correctly determine respectively 98.0% and 98.7% of the desired windows, when compared with 29,820 inspirations and 29,819 expirations detected by a human expert, obtained from three entire-night recordings. Postprocessing allowed determination of inspiration and expiration onsets with a mean difference with respect to the same human expert of (mean +/- SD) 34 +/- 71 ms for inspiration and 5 +/- 46 ms for expiration. The method proved to be effective in detecting the onset of inspiration and expiration in full night continuous recordings. A comparison of five human experts performing the same classification task yielded that the automated algorithm was undifferentiable from these human experts, falling within the distribution of human expert results. Besides being applicable to adult respiratory volume data, the presented algorithm was also successfully applied to infant sleep data, consisting of uncalibrated rib cage and abdominal movement recordings. A comparison with two previously published algorithms for breath detection in respiratory volume signal shows that the presented algorithm has a higher specificity, while presenting similar or higher positive predictive values.

Adult↗

PubMed2002

Classification of malignant and benign tumors using boundary characteristics in breast ultrasonograms.

We evaluated various spiculate and jagged margin shape features. These are known to be malignant characteristics in breast sonograms. A total of 79 breast ultrasonograms (60 benign, 19 malignant) containing solid breast nodules were evaluated. To determine the boundary of lesions, Markov random field segmentation was used. Our goal was to classify benign and malignant lesions on the breast sonogram. Our algorithm consisted of two steps: segmentation and classification. In the first step, a breast sonogram was segmented using low resolution and Gaussian-Markov random field. The fuzzy clustering method algorithm was then applied to the preprocessed image to initialize the segmentation. Next, to discriminate benign and malignant tumors three types of lesion characteristics were investigated: jag count, compactness, and acutance. Jag count was calculated based on the derivative of curvature, acutance was defined as gray-level variations across the lesion boundary, and compactness was defined as the ratio of boundary complexity to the enclosed area. Sensitivity of the three boundary features (jag count, compactness, and acutance) was 95.1, 94.1, and 81.1%, respectively, and their specificities were 97.2, 92.0, and 78%, respectively. The jag count performed best among the three boundary features. Our results indicate that computerized analysis of boundary characteristics can be an effective method for classifying solid breast nodules in ultrasonograms as malignant or benign. We found that curvature analysis was the best shape features. The curvature method classifies better than the compactness or acutance methods.

Breast Neoplasms↗

PubMed2001

A rapid classification protocol for the CATH Domain Database to support structural genomics.

In order to support the structural genomic initiatives, both by rapidly classifying newly determined structures and by suggesting suitable targets for structure determination, we have recently developed several new protocols for classifying structures in the CATH domain database (http://www.biochem.ucl.ac.uk/bsm/cath). These aim to increase the speed of classification of new structures using fast algorithms for structure comparison (GRATH) and to improve the sensitivity in recognising distant structural relatives by incorporating sequence information from relatives in the genomes (DomainFinder). In order to ensure the integrity of the database given the expected increase in data, the CATH Protein Family Database (CATH-PFDB), which currently includes 25,320 structural domains and a further 160,000 sequence relatives has now been installed in a relational ORACLE database. This was essential for developing more rigorous validation procedures and for allowing efficient querying of the database, particularly for genome analysis. The associated Dictionary of Homologous Superfamilies [Bray,J.E., Todd,A.E., Pearl,F.M.G., Thornton,J.M. and Orengo,C.A. (2000) Protein Eng., 13, 153-165], which provides multiple structural alignments and functional information to assist in assigning new relatives, has also been expanded recently and now includes information for 903 homologous superfamilies. In order to improve coverage of known structures, preliminary classification levels are now provided for new structures at interim stages in the classification protocol. Since a large proportion of new structures can be rapidly classified using profile-based sequence analysis [e.g. PSI-BLAST: Altschul,S.F., Madden,T.L., Schaffer,A.A., Zhang,J., Zhang,Z., Miller,W. and Lipman,D.J. (1997) Nucleic Acids Res., 25, 3389-3402], this provides preliminary classification for easily recognisable homologues, which in the latest release of CATH (version 1.7) represented nearly three-quarters of the non-identical structures.

Computational Biology↗

PubMedDate not supplied

Image segmentation in digital mammography: comparison of local thresholding and region growing algorithms.

Local thresholding and region-growing algorithms are developed and applied to digitized mammograms to quantify the parenchymal densities. The algorithms are first evaluated and optimized on phantom images reflecting varying image contrast, X-ray exposure conditions, and time-related changes. The difference between the segmentation results of the two techniques is less than 6% on the phantom images and 11% on the mammograms. The agreement between the computerized procedures and a manual one is in the range of 74-98%, depending on the breast parenchymal pattern and segmentation algorithm. The results show that computerized parenchymal classification of digitized mammograms is possible and independent of exposure.

Algorithms↗

PubMed2002

Single-trial classification of elementary comparison processes on the basis of instantaneous EEG and MEG coherences.

The present study is focused on the evidence of possible single-trial EEG/MEG analysis of information processing. The discrimination between thinking modalities of concept activation and pattern comparison for single tasks of elementary comparison procedures is investigated. A neural network classifier with backpropagation learning algorithm is used. The input vector is constructed by parameters of instantaneous coherence (13-20 Hz) between several channel pairs of the EEG and/or of the MEG. Thereby, the strength of synchronization and the time location of synchronization phenomena are taken into consideration. The combination of EEG and MEG coherence parameters led to a classification accuracy of 85-94% for single subjects. Generally, results reached by neural network classifier show a better generalization than linear discriminant analysis.

Adult↗

PubMed1991

Automatic segmentation and classification of ionic-channel signals.

Identification of ionic-channel types and their selectivity depends critically on the open channel current that can be resolved. In this paper, an automatic channel detection algorithm is proposed that is based on sequential minimization of an index which is usually used in cluster analysis. The algorithm consists of two stages, namely segmentation and classification. In the first stage, the signal samples are segmented based on the assumption that the samples in each segment should be sequentially connected. In the second stage, the resultant segments are classified with no regard to their connectivities. Results on synthetic and real channel currents are very encouraging and they suggest that this algorithm will substantially increase the productivity of many laboratories involved in ionic-channel research.

Algorithms↗

PubMed2003

Classification of transcranial Doppler signals using artificial neural network.

Transcranial Doppler signals, recorded from the temporal region of brain on 110 patients were transferred to a personal computer by using a 16-bit sound card. The fast Fourier transform (FFT) method was applied to the recorded signal from each patient. Since FFT method inherently can not offer a good spectral resolution at jet blood flows, it sometimes causes wrong interpretation of transcranial Doppler signals. To do a correct and rapid diagnosis, transcranial Doppler blood flow signals were statistically arranged so that they were classified in artificial neural network. Back propagation neural network and self-organization map algorithms of artificial neural network were used for training, whereas momentum and delta-bar-delta algorithms were used for learning. The results of these algorithms were compared in the case of classification and learning.

Adult↗

PubMed1998

Visual classification of medical data using MLP mapping.

In this work we discuss the design of a novel non-linear mapping method for visual classification based on multilayer perceptrons (MLP) and assigned class target values. In training the perceptron, one or more target output values for each class in a 2-dimensional space are used. In other words, class membership information is interpreted visually as closeness to target values in a 2D feature space. This mapping is obtained by training the multilayer perceptron (MLP) using class membership information, input data and judiciously chosen target values. Weights are estimated in such a way that each training feature of the corresponding class is forced to be mapped onto the corresponding 2-dimensional target value.

Algorithms↗

PubMed2006

A wavelet based method for automatic detection of slow eye movements: a pilot study.

Electro-oculographic (EOG) activity during the wake-sleep transition is characterized by the appearance of slow eye movements (SEM). The present work describes an algorithm for the automatic localisation of SEM events from EOG recordings. The algorithm is based on a wavelet multiresolution analysis of the difference between right and left EOG tracings, and includes three main steps: (i) wavelet decomposition down to 10 detail levels (i.e., 10 scales), using Daubechies order 4 wavelet; (ii) computation of energy in 0.5s time steps at any level of decomposition; (iii) construction of a non-linear discriminant function expressing the relative energy of high-scale details to both high- and low-scale details. The main assumption is that the value of the discriminant function increases above a given threshold during SEM episodes due to energy redistribution toward higher scales. Ten EOG recordings from ten male patients with obstructive sleep apnea syndrome were used. All tracings included a period from pre-sleep wakefulness to stage 2 sleep. Two experts inspected the tracings separately to score SEMs. A reference set of SEM (gold standard) were obtained by joint examination by both experts. Parameters of the discriminant function were assigned on three tracings (design set) to minimize the disagreement between the system classification and classification by the two experts; the algorithm was then tested on the remaining seven tracings (test set). Results show that the agreement between the algorithm and the gold standard was 80.44+/-4.09%, the sensitivity of the algorithm was 67.2+/-7.37% and the selectivity 83.93+/-8.65%. However, most errors were not caused by an inability of the system to detect intervals with SEM activity against NON-SEM intervals, but were due to a different localisation of the beginning and end of some SEM episodes. The proposed method may be a valuable tool for computerized EOG analysis.

Adult↗

PubMed2004

On the structure and initial parameter identification of Gaussian RBF networks.

We consider the efficient initialization of structure and parameters of generalized Gaussian radial basis function (RBF) networks using fuzzy decision trees generated by fuzzy ID3 like induction algorithms. The initialization scheme is based on the proposed functional equivalence property of fuzzy decision trees and generalized Gaussian RBF networks. The resulting RBF network is compact, easy to induce, comprehensible, and has acceptable classification accuracy with stochastic gradient descent learning algorithm.

Algorithms↗

PubMed1994

HYCONES II: a tool to build hybrid connectionist expert systems.

This paper describes HYCONES II--a tool to enable the construction of hybrid connectionist expert systems to solve classification problems. HYCONES II offers to the knowledge engineer a hybrid knowledge base that integrates frames with three different neural network models: the combinatorial neural model--CNM, the Fuzzy ARTMAP and the Semantic ART--SMART models. The latter is a new model, introduced by this paper, based on a combination of the two previous models. The validation section compares the performance of these three neural models to solve diagnostic problems in two medical domains. This paper also presents HYCONES II knowledge representation features, built in the symbolic component of its hybrid knowledge-base, to deal and represent fuzzy medical variables. Finally, the present status and future developments of the project are presented.

Algorithms↗

PubMed2007

Assessment of quantitative FDG PET data in primary colorectal tumours: which parameters are important with respect to tumour detection?

PURPOSE: The impact of quantitative parameters on the differentiation of primary colorectal tumours from normal colon tissue was assessed. Dynamic PET data (DPET) were acquired, and compartment and non-compartment modelling applied. The discriminant power of single parameters and the combination of PET parameters was assessed. All lesions were confirmed by histology. METHODS: FDG DPET studies were acquired in 22 patients with colorectal tumours prior to surgery. Five of these patients also had liver metastases at the time of the PET study. The SUV 56-60 min p.i. was included in the evaluation. A two-tissue compartment model was applied and the parameters k1-k4 as well as the fractional blood volume (VB) were obtained. The FDG influx was calculated from the compartment data. Non-compartment modelling was used to calculate the fractal dimension (FD) of the time-activity data. RESULTS: FD, SUV, influx and k3 were the most important single parameters for lesion differentiation. The highest accuracy was achieved for FD (88.78%). The overall tracer uptake was mainly dependent on k3 and not on k1 or VB. The support vector machines (SVM) algorithm was used to predict the classification based on the combination of individual PET parameters. The overall accuracy was 97.3%, with only one false positive case and no false negative results. The analysis of the subgroup of five patients with primary tumours and synchronous metastases revealed no significant differences for the individual PET parameters. However, VB tended to be lower while k1 and k2 were higher in patients with synchronous metastases. The SVM classification analysis predicted the presence of metastases based on the PET data of the primary tumour in three of five patients. CONCLUSION: Quantitative FDG PET studies provide very accurate data for the differentiation of primary colorectal tumours from normal tissue. The use of quantitative data has the advantage that the detection of a colorectal tumour is not primarily dependent on the individual assessment and experience of the physician evaluating the FDG PET data only visually. The results suggest that the presence of metastatic lesions may be predicted by analysis of the dynamic PET data of the corresponding primary tumour. Further studies are needed to assess this aspect in detail.

Algorithms↗