Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “unsupervised clustering”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

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

At least 415 records · Page 23Linked to original sources
PubMed2001

An investigation of competitive learning for autonomous cluster identification in embedded systems.

Robust signal processing for embedded systems requires the effective identification and representation of features within raw sensory data. This task is inherently difficult due to unavoidable long-term changes in the sensory systems and/or the sensed environment. In this paper we explore four variations of competitive learning and examine their suitability as an unsupervised technique for the automated identification of data clusters within a given input space. The relative performance of the four techniques is evaluated through their ability to effectively represent the structure underlying artificial and real-world data distributions. As a result of this study it was found that frequency sensitive competitive learning provides both reliable and efficient solutions to complex data distributions. As well, frequency sensitive and soft competitive learning are shown to exhibit properties which may permit the evolution of an appropriate network structure through the use of growing or pruning procedures.

Cluster Analysis↗
PubMed2006

Characterization of surface marine sediments from Ría de Arousa estuary according to extractable humic matter content.

Surface marine sediments from Ría de Arousa estuary were analyzed for humic and fulvic acids by UV-visible spectrometry and have been characterized using elemental analysis (carbon, hydrogen and nitrogen elemental composition) and spectrometric data (A2/A4 ratio, absorbancies at 270 and 407 nm and E4/E6 ratio, absorbancies at 465 and 665 nm). These variables have been used as discriminating factors to distinguish of marine and terrestrial origin of humic and fulvic acids in Ría de Arousa surface marine sediments. Principal component analysis, PCA, and cluster analysis, CA, have been used as unsupervised pattern recognition procedures. The half-range central value transformation was used as data pre-treatment to homogenize data. After a Varimax rotation, PCA applied to humic acid data has reveled that spectrometric A2/A4 and E4/E6 ratios are the main dominating features in the first principal component (48.6% of total variance), the humic acid content is the feature with the highest weight in the second principal component (22.9% of the total variability) and the carbon elemental composition domain in the third principal component (13.3% of total variance). Results from PCA have revealed that surface sediments collected at inner-left part of the estuary and at the mouth of the river Ulla belong to the same group. Similarly, PCA has shown that surface sediments from the right mouth of the estuary form a compact group. Taking in account the water circulation in Ría de Arousa estuary, these findings mean that the organic matter in surface sediments from the inner-left part of the estuary derived mainly from terrestrial organic matter while the organic matter in surface sediments from the right mouth of the estuary is mainly derived from marine sources. Finally, it must be noticed that any classification of surface sediments was assessed when applying of PCA and CA from fulvic acids data.

Carbon↗
PubMed2004

Cell protection, resistance and invasiveness of two malignant mesotheliomas as assessed by 10K-microarray.

Malignant pleural mesothelioma (MPM) is an aggressive serosal tumor, strongly associated with former exposure to asbestos fibers and for which there is currently no effective treatment available. In human, MPM is characterized by a high local invasiveness, poor prognosis and therapeutic outcomes. In order to assess molecular changes that specify this phenotype, we performed a global gene expression profiling of human MPM. Using a 10,000-element microarray, we analyzed mRNA relative gene expression levels by comparing a mesothelioma cell line to either a pleural cell line or tumor specimens. To analyze these gene expression data, we used various bioinformatics softwares. Hierarchical clustering methods were used to group genes and samples with similar expression in an unsupervised mode. Genes of known function were further sorted by enzyme, function and pathway clusters using a supervised software (IncyteGenomics). Taken together, these data defined a molecular fingerprint of human MPM with more than 700 up- or down-regulated genes related to several traits of the malignant phenotype, specially associated with MPM invasiveness, protection and resistance to anticancer defenses. This portrait is meaningful in disease classification and management, and relevant in finding new specific markers of MPM. These molecular markers should improve the accuracy of mesothelioma diagnosis, prognosis and therapy.

Biomarkers, Tumor↗
PubMed2005

Classification of amino acids induced by their associated matrices.

In this paper we carry out an analysis of different types of potential and substitution matrices for amino acids, oriented to give a classification of the latter. The cluster decomposition is obtained, in a fully unsupervised way, from the subdominant ultrametric associated to the distance between amino acids induced by the corresponding matrix. In the comparative study, by looking at the classifications obtained from diverse matrices, we can get information on how they account for the different chemical-physical properties of the amino acids.

Amino Acids↗
PubMed2001

Automatic detection and classification of abnormalities for artificial hearts using a hierarchical self-organizing map.

A hierarchical self-organizing map (SOM) has been developed for automatic detection and classification of abnormalities for artificial hearts. The hierarchical SOM has been applied to the monitoring and analysis of an aortic pressure (AoP) signal measured from an adult goat equipped with a total artificial heart. The architecture of the network actually consists of 2 different SOMs. The first SOM clusters the AoP beat patterns in an unsupervised way. Afterward, the outputs of the first SOM combined with the original time-domain features of beat-to-beat data are fed to the second SOM for final classification. Each input vector of the second SOM is associated with a class vector. This class vector is assigned to every node in the second map as an output weight and learned according to Kohonen's learning rule. Some experimental results revealed that a certain abnormality caused by breakage of sensors could be identified and detected correctly and that the change in the state of the circulatory system could be recognized and predicted to some extent.

Animals↗
PubMed1998

Rapid identification of urinary tract infection bacteria using hyperspectral whole-organism fingerprinting and artificial neural networks.

Three rapid spectroscopic approaches for whole-organism fingerprinting-pyrolysis mass spectrometry (PyMS), Fourier transform infra-red spectroscopy (FT-IR) and dispersive Raman microscopy--were used to analyse a group of 59 clinical bacterial isolates associated with urinary tract infection. Direct visual analysis of these spectra was not possible, highlighting the need to use methods to reduce the dimensionality of these hyperspectral data. The unsupervised methods of discriminant function and hierarchical cluster analyses were employed to group these organisms based on their spectral fingerprints, but none produced wholly satisfactory groupings which were characteristic for each of the five bacterial types. In contrast, for PyMS and FT-IR, the artificial neural network (ANN) approaches exploiting multi-layer perceptrons or radial basis functions could be trained with representative spectra of the five bacterial groups so that isolates from clinical bacteriuria in an independent unseen test set could be correctly identified. Comparable ANNs trained with Raman spectra correctly identified some 80% of the same test set. PyMS and FT-IR have often been exploited within microbial systematics, but these are believed to be the first published data showing the ability of dispersive Raman microscopy to discriminate clinically significant intact bacterial species. These results demonstrate that modern analytical spectroscopies of high intrinsic dimensionality can provide rapid accurate microbial characterization techniques, but only when combined with appropriate chemometrics.

Bacterial Typing Techniques↗
PubMed2004

Karyometry detects subvisual differences in chromatin organisation state between non-recurrent and recurrent papillary urothelial neoplasms of low malignant potential.

AIM: To analyse nuclear chromatin texture in non-recurrent and recurrent papillary urothelial neoplasms of low malignant potential (PUNLMPs). MATERIALS: Ninety three karyometric features were analysed on haematoxylin and eosin stained sections from 20 PUNLMP cases: 10 from patients with a solitary PUNLMP lesion, who were disease free during at least eight years' follow up, and 10 from patients with unifocal PUNLMP, one or more recurrences being seen during follow up. RESULTS: Kruskal-Wallis analysis was used to search for features showing significant differences between recurrent and non-recurrent cases. Significance was better than p<0.005 for more than 20 features. Based on significance, six texture features were selected for discriminant analysis. Stepwise linear discriminant analysis reduced Wilk's lambda to 0.87, indicating a highly significant difference between the two multivariate data sets, but only modest ability to discriminate (70% correct case classification). A box sequential classifier was used based on data derived from discriminant analysis. The classifier took three classification steps and classified 19 of the 20 cases correctly (95% correct case classification). To determine whether significant case grouping could also be obtained based on an objective criterion, the merged data sets of non-recurrent and recurrent cases were submitted to the unsupervised learning algorithm P-index. Two clusters were formed with significant differences. The subsequent application of a Cooley/Lohnes classifier resulted in an overall correct case classification rate of 85%. CONCLUSIONS: Karyometry and multivariate analyses detect subvisual differences in chromatin organisation state between non-recurrent and recurrent PUNLMPs, thus allowing identification of lesions that do or do not recur.

Algorithms↗
PubMed1999

Unsupervised classification of Space Acceleration Measurement System (SAMS) data using ART2-A.

The Space Acceleration Measurement System (SAMS) has been developed by NASA to monitor the microgravity acceleration environment aboard the space shuttle. The amount of data collected by a SAMS unit during a shuttle mission is in the several gigabytes range. Adaptive Resonance Theory 2-A (ART2-A), an unsupervised neural network, has been used to cluster these data and to develop cause and effect relationships among disturbances and the acceleration environment. Using input patterns formed on the basis of power spectral densities (psd), data collected from two missions, STS-050 and STS-057, have been clustered.

Acceleration↗
PubMed2003

Molecular classification of cancer: unsupervised self-organizing map analysis of gene expression microarray data.

An unsupervised self-organizing map-based clustering strategy has been developed to classify tissue samples from an oligonucleotide microarray patient database. Our method is based on the likelihood that a test data vector may have a gene expression fingerprint that is shared by more than one tumor class and as such can identify datasets that cannot be unequivocally assigned to a single tumor class. Our self-organizing map analysis completely separated the tumor from the normal expression datasets. Within the 14 different tumor types, classification accuracies on the order of approximately 80% correct were achieved. Nearly perfect classifications were found for leukemia, central nervous system, melanoma, uterine, and lymphoma tumor types, with very poor classifications found for colorectal, ovarian, breast, and lung tumors. Classification results were further analyzed to identify sets of differentially expressed genes between tumor and normal gene expressions and among each tumor class. Within the total pool of 1139 genes most differentially expressed in this dataset, subsets were found that could be vetted according to previously published literature sources to be specific tumor markers. Attempts to classify gene expression datasets from other sources found a wide range of classification accuracies. Discussions about the utility of this method and the quality of data needed for accurate tumor classifications are provided.

Algorithms↗
PubMed2006

Characterization of counterfeit artesunate antimalarial tablets from southeast Asia.

In southeast Asia, the widespread high prevalence of counterfeits tablets of the vital antimalarial artesunate is of great public health concern. To assess the seriousness of this problem, we quantified the amount of active ingredient present in artesunate tablets by liquid chromatography coupled to mass spectrometry. This method, in conjunction with analysis of the packaging, classified tablets as genuine, substandard, or fake and validated results of the colorimetric Fast Red TR test. Eight (35%) of 23 fake artesunate samples contained the wrong active ingredients, which were identified as different erythromycins and paracetamol. Raman spectroscopy identified calcium carbonate as an excipient in 9 (39%) of 23 fake samples. Multivariate unsupervised pattern recognition results indicated two major clusters of artesunate counterfeits, those with counterfeit foil stickers and containing calcium carbonate, erythromycin, and paracetamol, and those with counterfeit holograms and containing starch but without evidence of erythromycin or paracetamol.

Antimalarials↗
PubMed2006

Bayesian model averaging of naive Bayes for clustering.

This paper considers a Bayesian model-averaging (MA) approach to learn an unsupervised naive Bayes classification model. By using the expectation model-averaging (EMA) algorithm, which is proposed in this paper, a unique naive Bayes model that approximates an MA over selective naive Bayes structures is obtained. This algorithm allows to obtain the parameters for the approximate MA clustering model in the same time complexity needed to learn the maximum-likelihood model with the expectation-maximization algorithm. On the other hand, the proposed method can also be regarded as an approach to an unsupervised feature subset selection due to the fact that the model obtained by the EMA algorithm incorporates information on how dependent every predictive variable is on the cluster variable.

Algorithms↗
PubMed1999

Automatic segmentation of dynamic neuroreceptor single-photon emission tomography images using fuzzy clustering.

The segmentation of medical images is one of the most important steps in the analysis and quantification of imaging data. However, partial volume artefacts make accurate tissue boundary definition difficult, particularly for images with lower resolution commonly used in nuclear medicine. In single-photon emission tomography (SPET) neuroreceptor studies, areas of specific binding are usually delineated by manually drawing regions of interest (ROIs), a time-consuming and subjective process. This paper applies the technique of fuzzy c-means clustering (FCM) to automatically segment dynamic neuroreceptor SPET images. Fuzzy clustering was tested using a realistic, computer-generated, dynamic SPET phantom derived from segmenting an MR image of an anthropomorphic brain phantom. Also, the utility of applying FCM to real clinical data was assessed by comparison against conventional ROI analysis of iodine-123 iodobenzamide (IBZM) binding to dopamine D2/D3 receptors in the brains of humans. In addition, a further test of the methodology was assessed by applying FCM segmentation to [123I]IDAM images (5-iodo-2-[[2-2-[(dimethylamino)methyl]phenyl]thio] benzyl alcohol) of serotonin transporters in non-human primates. In the simulated dynamic SPET phantom, over a wide range of counts and ratios of specific binding to background, FCM correlated very strongly with the true counts (correlation coefficient r2>0.99, P<0.0001). Similarly, FCM gave segmentation of the [123I]IBZM data comparable with manual ROI analysis, with the binding ratios derived from both methods significantly correlated (r2=0.83, P<0.0001). Fuzzy clustering is a powerful tool for the automatic, unsupervised segmentation of dynamic neuroreceptor SPET images. Where other automated techniques fail completely, and manual ROI definition would be highly subjective, FCM is capable of segmenting noisy images in a robust and repeatable manner.

Animals↗
PubMed1998

Unsupervised classification of neural spikes with a hybrid multilayer artificial neural network.

The understanding of the brain structure and function and its computational style is one of the biggest challenges both in Neuroscience and Neural Computation. In order to reach this and to test the predictions of neural network modeling, it is necessary to observe the activity of neural populations. In this paper we propose a hybrid modular computational system for the spike classification of multiunits recordings. It works with no knowledge about the waveform, and it consists of two moduli: a Preprocessing (Segmentation) module, which performs the detection and centering of spike vectors using programmed computation; and a Processing (Classification) module, which implements the general approach of neural classification: feature extraction, clustering and discrimination, by means of a hybrid unsupervised multilayer artificial neural network (HUMANN). The operations of this artificial neural network on the spike vectors are: (i) compression with a Sanger Layer from 70 points vector to five principal component vector; (ii) their waveform is analyzed by a Kohonen layer; (iii) the electrical noise and overlapping spikes are rejected by a previously unreported artificial neural network named Tolerance layer; and (iv) finally the spikes are labeled into spike classes by a Labeling layer. Each layer of the system has a specific unsupervised learning rule that progressively modifies itself until the performance of the layer has been automatically optimized. The procedure showed a high sensitivity and specificity also when working with signals containing four spike types.

Animals↗
PubMed2000

Clustering ECG complexes using hermite functions and self-organizing maps.

An integrated method for clustering of QRS complexes is presented which includes basis function representation and self-organizing neural networks (NN's). Each QRS complex is decomposed into Hermite basis functions and the resulting coefficients and width parameter are used to represent the complex. By means of this representation, unsupervised self-organizing NN's are employed to cluster the data into 25 groups. Using the MIT-BIH arrhythmia database, the resulting clusters are found to exhibit a very low degree of misclassification (1.5%). The integrated method outperforms, on the MIT-BIH database, both a published supervised learning method as well as a conventional template cross-correlation clustering method.

Algorithms↗
PubMed1997

Computer-assisted analysis of medulloblastoma. A cytologic study.

OBJECTIVE: To explore data from a set of cases of medulloblastoma to see whether quantitative image analysis might suggest evidence for the existence of lower and higher grade lesions. STUDY DESIGN: Fourteen consecutive cases of medulloblastoma were obtained. Smears were stained with toluidine blue. For each case, 50 nuclei were measured and a number of densitometric features extracted. RESULTS: The existence of two subgroups of cases, identified as lower and higher grade groups, was suggested by a plot of the total optical density versus nuclear area. Two nuclear texture features--the number of pixels with the same optical density value occurring consecutively in the nucleus and the proportion of pixels in the high optical density range--divided the cases into the same subgroups. The use of a clustering algorithm established two clusters that corresponded to that subgrouping except for one case. Discriminant analysis gave an identical classification, with the misplaced case having a borderline discriminant function score. An unsupervised learning algorithm based on an adaptive distance metric formed two clusters and assigned the borderline case to the low grade subgroup. The grouping obtained by quantitative analysis was only partly related to the grade of nuclear atypia subjectively evaluated. CONCLUSION: In our series of medulloblastomas, quantitative analysis provided a means of detecting differences in the nuclear size and texture that allowed the classification of cases into two subgroups.

Adult↗
PubMed1996

Rapid identification of Streptococcus and Enterococcus species using diffuse reflectance-absorbance Fourier transform infrared spectroscopy and artificial neural networks.

Diffuse reflectance-absorbance Fourier transform infrared spectroscopy (FT-IR) was used to analyse 19 hospital isolates which had been identified by conventional means to one Enterococcus faecalis, E. faecium, Streptococcus bovis, S. mitis, S. pneumoniae, or S. pyogenes. Principal components analysis of the FT-IR spectra showed that this 'unsupervised' learning method failed to form six separable clusters (one of each species) and thus could not be used to identify these bacteria base on their FT-IR spectra. By contrast, artificial neural networks (ANNs) could be trained by 'supervised' learning (using the back-propagation algorithm) with the principal components scores of derivatised spectra to recognise the strains from their FT-IR spectra. These results demonstrate that the combination of FT-IR and ANNs provides a rapid, novel and accurate bacterial identification technique.

Algorithms↗
PubMed2002

PCBs in dungeness crab reflect distinct source fingerprints among harbor/industrial sites in British Columbia.

Dungeness crab (Cancer magister) samples were collected from various pulp mill and principal harbor sites on the West Coast of Canada. Full congener PCB analysis was performed on several composite and single hepatopancreas samples from each site, and the spatial variability of PCB patterns was explored. A recently developed direct mixing model (DMM) which relies on iteration of representative Aroclor end-members was used to make source predictions based on congener-specific PCB data from biota. Additionally, factor analysis and principal component analysis (FA/PCA) were applied to examine the intersite variability for potential PCB-source patterns. This unsupervised exploratory analysis (i.e., FA/PCA) revealed three distinct clusters of variables containing either low (di-tetra), moderate (penta-hexa), and high (hepta-nona) levels of chlorination, which were related to common Aroclor mixtures (e.g., A1242, A1248, A1254, and A1260). Overall, the PCA scores for each site qualitatively agreed with the source predictions provided by the DMM, and distinct source compositions were predicted for various sites examined.

Animals↗
PubMedDate not supplied

Data-driven analysis approach for biomarker discovery using molecular-profiling technologies.

High-throughput molecular-profiling technologies provide rapid, efficient and systematic approaches to search for biomarkers. Supervised learning algorithms are naturally suited to analyse a large amount of data generated using these technologies in biomarker discovery efforts. The study demonstrates with two examples a data-driven analysis approach to analysis of large complicated datasets collected in high-throughput technologies in the context of biomarker discovery. The approach consists of two analytic steps: an initial unsupervised analysis to obtain accurate knowledge about sample clustering, followed by a second supervised analysis to identify a small set of putative biomarkers for further experimental characterization. By comparing the most widely applied clustering algorithms using a leukaemia DNA microarray dataset, it was established that principal component analysis-assisted projections of samples from a high-dimensional molecular feature space into a few low dimensional subspaces provides a more effective and accurate way to explore visually and identify data structures that confirm intended experimental effects based on expected group membership. A supervised analysis method, shrunken centroid algorithm, was chosen to take knowledge of sample clustering gained or confirmed by the first step of the analysis to identify a small set of molecules as candidate biomarkers for further experimentation. The approach was applied to two molecular-profiling studies. In the first study, PCA-assisted analysis of DNA microarray data revealed that discrete data structures exist in rat liver gene expression and correlated with blood clinical chemistry and liver pathological damage in response to a chemical toxicant diethylhexylphthalate, a peroxisome-proliferator-activator receptor agonist. Sixteen genes were then identified by shrunken centroid algorithm as the best candidate biomarkers for liver damage. Functional annotations of these genes revealed roles in acute phase response, lipid and fatty acid metabolism and they are functionally relevant to the observed toxicities. In the second study, 26 urine ions identified from a GC/MS spectrum, two of which were glucose fragment ions included as positive controls, showed robust changes with the development of diabetes in Zucker diabetic fatty rats. Further experiments are needed to define their chemical identities and establish functional relevancy to disease development.

Algorithms↗