Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Classification Algorithms”

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 163 records · Page 9Linked to original sources
PubMed2006

Kalman filtering for disease-state estimation from microarray data.

MOTIVATION: In this paper, we propose using the Kalman filter (KF) as a pre-processing step in microarray-based molecular diagnosis. Incorporating the expression covariance between genes is important in such classification problems, since this represents the functional relationships that govern tissue state. Failing to fulfil such requirements may result in biologically implausible class prediction models. Here, we show that employing the KF to remove noise (while retaining meaningful covariance and thus being able to estimate the underlying biological state from microarray measurements) yields linearly separable data suitable for most classification algorithms. RESULTS: We demonstrate the utility and performance of the KF as a robust disease-state estimator on publicly available binary and multi-class microarray datasets in combination with the most widely used classification methods to date. Moreover, using popular graphical representation schemes we show that our filtered datasets also have an improved visualization capability.

Algorithms↗
PubMedDate not supplied

Diagnosing breast cancer based on support vector machines.

The Support Vector Machine (SVM) classification algorithm, recently developed from the machine learning community, was used to diagnose breast cancer. At the same time, the SVM was compared to several machine learning techniques currently used in this field. The classification task involves predicting the state of diseases, using data obtained from the UCI machine learning repository. SVM outperformed k-means cluster and two artificial neural networks on the whole. It can be concluded that nine samples could be mislabeled from the comparison of several machine learning techniques.

Algorithms↗
PubMed2006

Automatic determination of acoustic plate source-detector separation from one waveform.

We discuss an automated procedure for determining the separation between a transient acoustic source and a detector on a plate. We use a time-estimation algorithm based on the assumption that the detected signal is represented by a small, finite number of discrete band-limited impulses. This is carried out using the MUSIC (MUltiple SIgnal Classification) algorithm in time-estimation mode to automatically estimate both the first arrival time of the lowest order antisymmetric (Ao) mode and the arrival time of the Rayleigh wave. Using the material and geometric properties of the plate and these two arrival times, we calculated the distance to the source. This technique allows the automatic determination of source-receiver separation from a single transient waveform.

Journal Article↗
PubMed2002

Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men.

The prostate-specific antigen test has been a major factor in increasing awareness and better patient management of prostate cancer (PCA), but its lack of specificity limits its use in diagnosis and makes for poor early detection of PCA. The objective of our studies is to identify better biomarkers for early detection of PCA using protein profiling technologies that can simultaneously resolve and analyze multiple proteins. Evaluating multiple proteins will be essential to establishing signature proteomic patterns that distinguish cancer from noncancer as well as identify all genetic subtypes of the cancer and their biological activity. In this study, we used a protein biochip surface enhanced laser desorption/ionization mass spectrometry approach coupled with an artificial intelligence learning algorithm to differentiate PCA from noncancer cohorts. Surface enhanced laser desorption/ionization mass spectrometry protein profiles of serum from 167 PCA patients, 77 patients with benign prostate hyperplasia, and 82 age-matched unaffected healthy men were used to train and develop a decision tree classification algorithm that used a nine-protein mass pattern that correctly classified 96% of the samples. A blinded test set, separated from the training set by a stratified random sampling before the analysis, was used to determine the sensitivity and specificity of the classification system. A sensitivity of 83%, a specificity of 97%, and a positive predictive value of 96% for the study population and 91% for the general population were obtained when comparing the PCA versus noncancer (benign prostate hyperplasia/healthy men) groups. This high-throughput proteomic classification system will provide a highly accurate and innovative approach for the early detection/diagnosis of PCA.

Aged↗
PubMed2000

Automated classification of human atrial fibrillation from intraatrial electrograms.

The assessment of the degree of organization and the classification of atrial fibrillation (AF) according to the types defined by Wells usually resorts to the visual inspection of bipolar intraatrial electrograms. The focus of this study was to test seven parameters aimed to quantify the degree of organization of the electrograms, and then to design a final classification scheme based on a multidimensional, minimum-distance analysis. The following parameters were tested: mean atrial period (AP) and its coefficient of variation (CV); number of points lying at the baseline (NO) and the Shannon entropy (EN) of the amplitude probability density function (APDF); depolarization width (F-WIDTH); and correlation waveform analysis (CWA) and electrogram bandwidth (BW). The signal database consisted in a set of 160 AF strips of Type I, II, and III AF, scored by an expert cardiologist (60 Type I, 40 Type II, 60 Type III) and further divided in a training set (60) and a test set (100). Strips were 6 seconds long and were recorded with 5-mm interspace bipolar catheters from electrically induced (n = 13) and chronic (n = 10) patients. A classification algorithm based on a minimum-distance (Mahalanobis distance) discriminant analysis was tested. Using a single parameter, the best discriminations were provided by NO, F-WIDTH, and CV. F-WIDTH was found strongly inversely correlated to NO (r = -0.90). Of all the two-parameter combinations, CV-NO provided the best classification: 92 of 100 segments were correctly classified with sensitivity > 90% and specificity > 92%. A further improvement was obtained by including BW as a third parameter (93/100 correctly classified). The use of more than three parameters not only failed to improve, but even degraded the classification.

Algorithms↗
PubMed1997

Prognostic factors: classification approaches in patients with lung cancer.

With the proliferation of potential prognostic factors for lung cancer, it is becoming increasingly more difficult to integrate the information provided by these factors into a single accurate prediction of clinical outcome. Here we reviewed five classification methods for their capabilities in classification of 200 patients with lung cancer into distinct prognostic groups using survival outcome as a criteria. The source of patient data for this study is a Lung Tumour Registry from Institute for Lung Diseases, University Clinical Hospital, Belgrade. Almost all developed classification algorithms determined prognostic groups according to biochemical tumour markers LDH and alkaline phosphatase, producing most significant split, instead of commonly used staging variables. The choice of which approach to use for a given classification problem depends not only on statistical properties of method, but also on medical considerations, such as whether more differential findings are given greater weight and the applicability of a classification rule.

Alkaline Phosphatase↗
PubMed2001

Drug design by machine learning: support vector machines for pharmaceutical data analysis.

We show that the support vector machine (SVM) classification algorithm, a recent development from the machine learning community, proves its potential for structure-activity relationship analysis. In a benchmark test, the SVM is compared to several machine learning techniques currently used in the field. The classification task involves predicting the inhibition of dihydrofolate reductase by pyrimidines, using data obtained from the UCI machine learning repository. Three artificial neural networks, a radial basis function network, and a C5.0 decision tree are all outperformed by the SVM. The SVM is significantly better than all of these, bar a manually capacity-controlled neural network, which takes considerably longer to train.

Algorithms↗
PubMedDate not supplied

A comparative study on feature selection methods for drug discovery.

Feature selection is frequently used as a preprocessing step to machine learning. The removal of irrelevant and redundant information often improves the performance of learning algorithms. This paper is a comparative study of feature selection in drug discovery. The focus is on aggressive dimensionality reduction. Five methods were evaluated, including information gain, mutual information, a chi2-test, odds ratio, and GSS coefficient. Two well-known classification algorithms, Naïve Bayesian and Support Vector Machine (SVM), were used to classify the chemical compounds. The results showed that Naïve Bayesian benefited significantly from the feature selection, while SVM performed better when all features were used. In this experiment, information gain and chi2-test were most effective feature selection methods. Using information gain with a Naïve Bayesian classifier, removal of up to 96% of the features yielded an improved classification accuracy measured by sensitivity. When information gain was used to select the features, SVM was much less sensitive to the reduction of feature space. The feature set size was reduced by 99%, while losing only a few percent in terms of sensitivity (from 58.7% to 52.5%) and specificity (from 98.4% to 97.2%). In contrast to information gain and chi2-test, mutual information had relatively poor performance due to its bias toward favoring rare features and its sensitivity to probability estimation errors.

Algorithms↗
PubMed2004

Automated segmentation of routinely hematoxylin-eosin-stained microscopic images by combining support vector machine clustering and active contour models.

OBJECTIVE: To develop a method for the automated segmentation of images of routinely hematoxylin-eosin (H-E)-stained microscopic sections to guarantee correct results in computer-assisted microscopy. STUDY DESIGN: Clinical material was composed 50 H-E-stained biopsies of astrocytomas and 50 H-E-stained biopsies of urinary bladder cancer. The basic idea was to use a support vector machine clustering (SVMC) algorithm to provide gross segmentation of regions holding nuclei and subsequently to refine nuclear boundary detection with active contours. The initialization coordinates of the active contour model were defined using a SVMC pixel-based classification algorithm that discriminated nuclear regions from the surrounding tissue. Starting from the boundaries of these regions, the snake fired and propagated until converging to nuclear boundaries. RESULTS: The method was validated for 2 different types of H-E-stained images. Results were evaluated by 2 histopathologists. On average, 94% of nuclei were correctly delineated. CONCLUSION: The proposed algorithm could be of value in computer-based systems for automated interpretation of microscopic images.

Algorithms↗
PubMed2003

Diagnostic potential of serum proteomic patterns in prostate cancer.

PURPOSE: The serum prostate specific antigen test has been widely used in the last decade as an effective screening tool for prostate cancer (CaP). However, the high false-positive rate of the serum prostate specific antigen test necessitates the development of more accurate diagnostic and prognostic biomarkers for CaP. Promising diagnostic potential of serum protein patterns detected by surface enhanced laser desorption/ionization time of flight mass spectrometry for CaP has recently been reported. Independent evaluation of this new technology is warranted to realize its translational utility. We determined whether serum protein profiling by surface enhanced laser desorption/ionization time of flight mass spectrometry and a decision tree algorithm classification system could accurately discriminate between patients with CaP and unaffected individuals. MATERIALS AND METHODS: Proteomic spectra of crude serum were generated using the Ciphergen ProteinChip System and pattern detection was performed using Biomarker Patterns Software (Ciphergen Biosystems, Inc., Fremont, California). A total of 106 patients with CaP and 56 controls were randomly allocated to a training set and a test set. The training set, which consisted of 44 patients with cancer and 30 controls, was used to build a decision tree algorithm. The test set, which consisted of 62 patients with cancer and 26 controls, was used in blinded fashion to validate the decision tree. RESULTS: Accuracy of classification using the test set was 67% and 42% for the weak cation exchange array and the copper metal affinity capture array, respectively. Combined spectral data from the weak cation exchange and copper metal affinity capture arrays generated an algorithm that achieved 85% sensitivity and 85% specificity for the detection of CaP. CONCLUSIONS: These preliminary findings support recent observations that complex protein profiles have promising potential for the early detection of CaP and warrant future studies with streamlined technology. Furthermore, the combined effect of using 2 array types can greatly enhance the ability of protein profile patterns, suggesting the potential usefulness of alternative approaches to evaluate this new emerging technology.

Algorithms↗
PubMedDate not supplied

Ad hoc classification of radiology reports.

OBJECTIVE: The task of ad hoc classification is to automatically place a large number of text documents into nonstandard categories that are determined by a user. The authors examine the use of statistical information retrieval techniques for ad hoc classification of dictated mammography reports. DESIGN: The authors' approach is the automated generation of a classification algorithm based on positive and negative evidence that is extracted from relevance-judged documents. Test documents are sorted into three conceptual bins: membership in a user-defined class, exclusion from the user-defined class, and uncertain. Documentation of absent findings through the use of negation and conjunction, a hallmark of interpretive test results, is managed by expansion and tokenization of these phrases. MEASUREMENTS: Classifier performance is evaluated using a single measure, the F measure, which provides a weighted combination of recall and precision of document sorting into true positive and true negative bins. RESULTS: Single terms are the most effective text feature in the classification profile, with some improvement provided by the addition of pairs of unordered terms to the profile. Excessive iterations of automated classifier enhancement degrade performance because of overtraining. Performance is best when the proportions of relevant and irrelevant documents in the training collection are close to equal. Special handling of negation phrases improves performance when the number of terms in the classification profile is limited. CONCLUSIONS: The ad hoc classifier system is a promising approach for the classification of large collections of medical documents. NegExpander can distinguish positive evidence from negative evidence when the negative evidence plays an important role in the classification.

Algorithms↗
PubMed2003

Molecular features of non-B, non-C hepatocellular carcinoma: a PCR-array gene expression profiling study.

BACKGROUND/AIMS: Hepatocellular carcinoma (HCC) usually develops following chronic liver inflammation caused by hepatitis C or B virus. Through expression profiling in a rare type of HCC, for which the causes are unknown, we sought to find key genes responsible for each step of hepatocarcinogenesis in the absence of viral influence. METHODS: We used 68 non-B, non-C liver tissues (20 HCC, 17 non-tumor, 31 normal liver) for expression profiling with PCR-array carrying 3072 genes known to be expressed in liver tissues. To select the differentially expressed genes, we performed random permutation testing. A weighted voting classification algorithm was used to confirm the reliability of gene selection. We then compared these genes with the results of previous expression profiling studies. RESULTS: A total of 220 differentially expressed genes were selected by random permutation tests. The classification accuracies using these genes were 91.8, 92.0 and 100.0% by a leave-one-out cross-validation, an additional PCR-array dataset and a Stanford DNA microarray dataset, respectively. By comparing our results with previous reports on virus-infected HCC, four genes (ALB, A2M, ECHS1 and IGFBP3) were commonly selected in some studies. CONCLUSIONS: The 220 differentially expressed genes selected by PCR-array are potentially responsible for hepatocarcinogenesis in the absence of viral influence.

Aged↗
PubMed2006

Treatment for Adolescents with Depression Study (TADS): safety results.

OBJECTIVE: To compare the rates of physical, psychiatric, and suicide-related events in adolescents with MDD treated with fluoxetine alone (FLX), cognitive-behavioral therapy (CBT), combination treatment (COMB), or placebo (PBO). METHOD: Safety assessments included adverse events (AEs) collected by spontaneous report, as well as systematic measures for specific physical and psychiatric symptoms. Suicidal ideation and suicidal behavior were systematically assessed by self- and clinician reports. Suicidal events were also reanalyzed by the Columbia Group and expert raters using the Columbia-Classification Algorithm for Suicidal Assessment used in the U.S. Food and Drug Administration reclassification effort. RESULTS: Depressed adolescents reported high rates of physical symptoms at baseline, which improved as depression improved. Sedation, insomnia, vomiting, and upper abdominal pain occurred in at least 2% of those treated with FLX and/or COMB and at twice the rate of placebo. The rate of psychiatric AEs was 11% in FLX, 5.6% in COMB, 4.5% in PBO, and 0.9% in CBT. Suicidal ideation improved overall, with greatest improvement in COMB. Twenty-four suicide-related events occurred during the 12-week period: 5 patients (4.7%) in COMB, 10 (9.2%) in FLX, 5 (4.5%) in CBT, and 3 (2.7%) in placebo. Statistically, only FLX had more suicide-related events than PBO (p =.0402, odds ratio (OR) = 3.7, 95% CI 1.00-63.7). Only five actual attempts occurred (2 COMB, 2 FLX, 1 CBT, 0 PBO). There were no suicide completions. CONCLUSIONS: Different methods for eliciting AEs produce different results. In general, as depression improves, physical complaints and suicidal ideation decrease in proportion to treatment benefit. In this study, psychiatric AEs and suicide-related events are more common in FLX-treated patients. COMB treatment may offer a more favorable safety profile than medication alone in adolescent depression.

Adolescent↗
PubMed2006

Finding diagnostic biomarkers in proteomic spectra.

In seeking to find diagnostic biomarkers in proteomic spectra, two significant problems arise. First, not only is there noise in the measured intensity at each m/z value, but there is also noise in the measured m/z value itself. Second, the potential for overfitting is severe: it is easy to find features in the spectra that accurately discriminate disease states but have no biological meaning. We address these problems by developing and testing a series of steps for pre-processing proteomic spectra and extracting putatively meaningful features before presentation to feature selection and classification algorithms. These steps include an HMM-based latent spectrum extraction algorithm for fusing the information from multiple replicate spectra obtained from a single tissue sample, a simple algorithm for baseline correction based on a segmented convex hull, a peak identification and quantification algorithm, and a peak registration algorithm to align peaks from multiple tissue samples into common peak registers. We apply these steps to MALDI spectral data collected from normal and tumor lung tissue samples, and then compare the performance of feature selection with FDR followed by classification with an SVM, versus joint feature selection and classification with Bayesian sparse multinomial logistic regression (SMLR). The SMLR approach outperformed FDR+SVM, but both were effective in achieving good diagnostic accuracy with a small number of features. Some of the selected features have previously been investigated as clinical markers for lung cancer diagnosis; some of the remaining features are excellent candidates for further research.

Algorithms↗
PubMed2004

Detection of differentially expressed proteins in early-stage melanoma patients using SELDI-TOF mass spectrometry.

Tumor progression is a dynamic sequence of events that involves specific protein changes. We hypothesized that Surface Enhanced Laser Desorption/Ionization (SELDI) mass spectrometric analysis of sera from patients with AJCC stage I and II melanoma with negative loco-regional lymph nodes could identify potential melanoma-associated protein biomarkers of disease recurrence. Serum specimens were collected from 49 patients who developed recurrence (n = 25) or remained free of recurrence (n = 24) without evidence of disease following complete resection (AJCC stage I and II). Follow-up was longer than 5 years. Serum proteins were denatured and applied onto two protein chip chemistry surfaces (weak cationic WCX2; metal-binding, IMAC3-Cu). SELDI ProteinChip mass spectrometry was then performed. SELDI data were analyzed, protein peak clustering and classification were performed, and a supervised classification algorithm was employed to classify the dataset. Multiple protein peaks ranging from 3.3 to 30 kDa were identified between patients with recurrence and those without recurrence, and the expression pattern differences of three proteins were used to generate the discriminating classification tree. The biomarkers were expressed with a high degree of reproducibility. In this early characterization study, melanoma recurrence was predicted with a sensitivity of 72% (18/25) and a specificity of 75% (18/24). This novel pilot study revealed three proteins that accurately identified patients who developed recurrence after curative resection of primary melanoma.

Adult↗
PubMed2006

Protein homology network families reveal step-wise diversification of Type III and Type IV secretion systems.

From the analysis of 251 prokaryotic genomes stored in public databases, the 761,260 deduced proteins were used to reconstruct a complete set of bacterial proteic families. Using the new Overlap algorithm, we have partitioned the Protein Homology Network (PHN), where the proteins are the nodes and the links represent homology relationships. The algorithm identifies the densely connected regions of the PHN that define the families of homologous proteins, here called PHN-Families, recognizing the phylogenetic relationships embedded in the network. By direct comparison with a manually curated dataset, we assessed that this classification algorithm generates data of quality similar to a human expert. Then, we explored the network to identify families involved in the assembly of Type III and Type IV secretion systems (T3SS and T4SS). We noticed that, beside a core of conserved functions (eight proteins for T3SS, seven for T4SS), a variable set of accessory components is always present (one to nine for T3SS, one to five for T4SS). Each member of the core corresponds to a single PHN-Family, while accessory proteins are distributed among different pure families. The PHN-Family classification suggests that T3SS and T4SS have been assembled through a step-wise, discontinuous process, by complementing the conserved core with subgroups of nonconserved proteins. Such genetic modules, independently recruited and probably tuned on specific effectors, contribute to the functional specialization of these organelles to different microenvironments.

Amino Acid Sequence↗
PubMed2005

Partially distributed representations of objects and faces in ventral temporal cortex.

Object and face representations in ventral temporal (VT) cortex were investigated by combining object confusability data from a computational model of object classification with neural response confusability data from a functional neuroimaging experiment. A pattern-based classification algorithm learned to categorize individual brain maps according to the object category being viewed by the subject. An identical algorithm learned to classify an image-based, view-dependent representation of the stimuli. High correlations were found between the confusability of object categories and the confusability of brain activity maps. This occurred even with the inclusion of multiple views of objects, and when the object classification model was tested with high spatial frequency "line drawings" of the stimuli. Consistent with a distributed representation of objects in VT cortex, the data indicate that object categories with shared image-based attributes have shared neural structure.

Algorithms↗
PubMed2005

Right- and left-sided colorectal cancers display distinct expression profiles and the anatomical stratification allows a high accuracy prediction of lymph node metastasis.

Accurate preoperative prediction of lymph node metastasis and degree of tumor invasion would facilitate an appropriate decision of the extent of surgical resection of cancers to reduce unnecessary complication or to minimize the risk of recurrence in patients. We analyzed gene expression profiles characteristic of the invasiveness of colorectal carcinoma in a total of 89 cases, using a cDNA array and pattern classification algorithms. We set binary classes for a panel of clinicopathologic parameters, each of which was divided at different levels for categories (discrete) or values (continuous). We searched an optimal combination of genes to discriminate the classes by using of a feature subset selection algorithm, which was applied to a set of genes preselected on the basis of statistical difference in expression (two-sided t test, P < or = 0.05). We used a sequential forward feature selection which additively searched a combination of genes, giving a minimal leave-one-out classification error rate of a k-nearest neighbor classifier. In the process of gene preselection, we found a remarkable difference in the expression pattern of genes according to the anatomical location of cancers. The difference was most prominent when the classes were set for cecum, ascending colon, transverse colon, and descending colon (CATD) versus sigmoid colon and rectum (SR). By stratifying these two locations, we were able to extract gene expression profiles characteristic of the classes of the presence versus absence of lymph node metastasis, lymphatic invasion, vascular invasion and degree of mural invasion, and pathological stages, with an accuracy of more than 90%. These results suggest that colorectal cancers harbor distinct molecular pathophysiological statuses according to their right-to-left locations, of which stratification is important for pattern classification of cDNA array data.

Adult↗