Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Feature selection”

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 1,081 records · Page 60Linked to original sources
PubMed1990

Karyometric marker features in normal-appearing glands adjacent to human colonic adenocarcinoma.

The expression of nuclear marker features in normal-appearing tissue adjacent to colonic adenocarcinoma was investigated. Formalin-fixed, paraffin-embedded tissue sections of colon from 9 patients with adenocarcinoma and from 9 normal controls were cut 4 microns thick, Feulgen stained, and measured by a cell image analysis system using a Matrox MVP-AT/NP imaging board. Thirty nuclei in the tumor region, 30 nuclei 2 mm into the histologically normal-appearing distal margin, and the same number at 5, 10, 20, and 50 mm into the margin were measured for each patient. An additional 30 nuclei were recorded from 9 patients each free from colonic disease. Nuclear features were selected to discriminate between tumor nuclei and nuclei from normal control subjects and between nuclei measured in the histologically normal-appearing margin next to the tumor and control nuclei. Eight micromorphometric measures were found to be statistically significantly different in nuclei measured in the margin site, including features describing staining density (total absorbance, average absorbance 20% below mean, average absorbance 20% above mean) chromatin texture (cooccurrence matrix, run length, and peripheral tendency) and nuclear area. The category differences are statistically highly significant.

Adenocarcinoma↗
PubMed2000

Reverberation at the Mid-Atlantic Ridge during the 1993 ARSRP experiment seen by R/V Alliance from 200-1400 Hz and some modeling inferences

In July 1993 SACLANTCEN participated in an experiment for the Acoustic Reverberation Special Research Program (ARSRP). The primary objective was to take high resolution measurements to shed light on the detailed physical processes dominating the low frequency scattering from rough topographic features and from deep sediment pond areas. A very detailed set of monostatic and bistatic scattering experiments were conducted just west of the Mid-Atlantic Ridge near 26 degrees N, 47 degrees W. Results from SACLANTCEN's R/V ALLIANCE are presented which show monostatic and bistatic scattering from selected bathymetric features using pulsed sources from 200-1400 Hz and SUS charges. Area water depths ranged from 3300 to greater than 5200 m. The receivers were a horizontal array of 128 elements spaced at 0.5, 1, and 2 m. Source depths varied from 120 to 1200 m and receiver depths ranged from 400 to 500 m. Received reverberation levels are presented as functions of the estimated scatterer position on area bathymetry maps. Results show the same high scattering areas from one seamount were observed in three distinct frequency bands centered at 227, 350, and 700 Hz. Even at 1350 Hz the correlation of scattering highlights with nearby bathymetric features is quite good. Comparisons of measured reverberation and modeled reverberation are presented which suggest that the dominant longer range scattering from the large features is driven by the steep slopes present above the critical depth but that small scale roughness is also essential for strong backscatter.

Journal Article↗
PubMed1976

Paraproteinemia: blood hyperviscosity and clinical manifestations.

Many of the clinical features of paraproteinemia result from impairment of blood flow through the vascular tree because of blood hyperviscosity. Studies were carried out in 65 patients with serum paraproteins (31 with IgG, 25 with IgM, and 9 with IgA) to examine the relationship between the blood viscosity and the frequency of selected clinical features. The blood and plasma viscosities were measured at low rates of shear. Blood hyperviscosity was present in 91% of the patients and plasma hyperviscosity in 75% of the patients. In each of the three immunoglobulin classes both the blood and plasma viscosities increased logarithmically with the paraprotein concentration being greatest in the case of IgM. In addition, the relationship between the hematocrit and the logarithm of blood viscosity tended to be linear at any given protein concentration. In patients with very high levels of paraprotein the blood viscosity was modified by low hematocrits; the latter was below 30 in 70% of patients in whom the concentration of paraprotein was above 4 g/100 ml. The prevalence of clinical complications involving the retinal circulation, the peripheral vascular system, and the central nervous system increased markedly with increasing blood viscosity, measured at 0.18 S-1. One or more of these regions was affected in greater than 80% of patients with blood viscosity above 60 centipoise and in less than 23% of patients with blood viscosity below 40 centipoise. These observations illustrate the complex relationship between blood viscosity, concentration of paraprotein, immunoglobulin class and hematocrit, and emphasize the importance of measuring the whole blood viscosity at low rates of shear in determining the risk of vascular complications.

Blood Viscosity↗
PubMed1999

Object selection based on oscillatory correlation.

One of the classical topics in neural networks is winner-take-all (WTA), which has been widely used in unsupervised (competitive) learning, cortical processing, and attentional control. Owing to global connectivity, WTA networks, however, do not encode spatial relations in the input, and thus cannot support sensory and perceptual processing where spatial relations are important. We propose a new architecture that maintains spatial relations between input features. This selection network builds on Locally Excitatory Globally Inhibitory Oscillator Networks (LEGION) dynamics and slow inhibition. In an input scene with many objects (patterns), the network selects the largest object. This system can be easily adjusted to select several largest objects, which then alternate in time. We analyze the speed of selection, and further show that a two-stage selection network gains efficiency by combining selection with parallel removal of noisy regions. The network is applied to select the most salient object in gray-level images. As a special case, the selection network without local excitation gives rise to a new form of oscillatory WTA.

Journal Article↗
PubMed1999

Three-dimensional anatomy setup verification by correlation of orthogonal portal images and digitally reconstructed radiographs.

A method for three-dimensional verification of anatomy setup that uses the correlation of portal images and reference megavoltage digitally reconstructed radiographs (MDRRs) is presented. Prior to a treatment, an orthogonal pair of portal images is acquired from which subimages containing anatomical features are selected. These subimages are consequently matched to MDRRs that represent different rotations of the anatomy around axes going through the treatment isocenter. The Pearson correlation coefficient is employed for the matching since it is invariant with respect to global scaling and shifting of the image intensities. Furthermore, it does not require feature extraction or point-pair identification. The greatest value of the correlation coefficient corresponds to the proper rotational alignment of the anatomy and the location of the correlation maximum in each view indicates the translational shifts of the anatomy. The mean accuracy of translation and rotation registrations tests were a fraction of a millimeter and a fraction of a degree, respectively, for MDRR-to-MDRR matching. For portal-to-MDRR matching, the mean translation registration error is on the order of 1 mm and the mean error in radial displacement is of the order of 2.7 mm.

Algorithms↗
PubMed2003

Diagnosing Hirschsprung's disease: increasing the odds of a positive rectal biopsy result.

BACKGROUND/PURPOSE: Rectal biopsies are performed as a definitive means of diagnosing Hirschsprung's Disease (HD) in children presenting with constipation. The authors hypothesized that key features in the history, physical examination, and radiographic evaluation would allow us to avoid unnecessary rectal biopsies. METHODS: A retrospective analysis was conducted on patients undergoing rectal biopsy between 1995 and 2001. Patients with HD were identified (n = 50), and a concurrent cohort of patients with idiopathic constipation (IC; n = 50) was selected. Pertinent features in patients with HD versus those with IC were cross tabulated using Pearson Chi2 testing (significance was P <.05). RESULTS: Sixty percent of patients with HD and 15% of patients with IC experienced onset of symptoms in the first week of life. HD patients more frequently experienced delayed passage of meconium (P <.05), abdominal distension (P <.05), vomiting (P <.05), and transition zone on contrast enema (P <.05). All patients with HD had one or more of these significant features. In contrast, only 64% of patients with IC had one or more of these features. The classic triad of symptoms (ie, delayed passage of meconium, vomiting, and abdominal distension) was present in 18%, and one or more of these symptoms was present in 98% of HD patients. In contrast, only 60% of patients with IC had a history of delayed passage of meconium, vomiting, or abdominal distension. CONCLUSIONS: A history of delayed passage of meconium, abdominal distension, vomiting or the results of a contrast enema identified all patients with HD and excluded HD in approximately 36% of patients with idiopathic constipation. The authors have shown that key features in a patient's history, physical examination, and radiologic evaluation can differentiate between HD and IC. In a child presenting with constipation and none of the above features, it is not necessary to perform a rectal biopsy to exclude HD.

Adolescent↗
PubMed1997

Selective IgG2 subclass deficiency--a marker for the syndrome of pre/postnatal growth retardation, developmental delay, hypotrophy of distal extremities, dental anomalies and eczema.

We report a third family with members displaying pre- and postnatal growth retardation, hypotonia, psychomotor retardation, small puffy hands and feet, dental anomalies and eczematous skin. The four affected members are all females born to unrelated parents consistent with the previously proposed autosomal recessive mode of inheritance. We report a further clinical feature of selective immunoglobulin IgG2 subclass deficiency which would explain some of the clinical findings and might provide an immunological marker for diagnostic confirmation of the syndrome.

Biomarkers↗
PubMed1988

Modeling of beta-adrenoceptors based on molecular electrostatic potential studies of agonists and antagonists.

The molecular electrostatic potential (MEP) of 32 beta-adrenoceptor ligands, mainly antagonists, was calculated by the STO-3G ab initio quantum mechanical method. The MEP of phenylethanolamines (PEAs) features a negative minimum in the meta region (designated M1) which is topographically equivalent to a minimum (designated M2) found in the vicinity of the aromatic ring in all (aryloxy)propanolamines (AOPAs). In these compounds, a second negative zone located beyond the meta position and designated M3 is found in all beta 1-selective antagonists and in some nonselective and beta 2-selective antagonists. The beta 1-selective antagonists feature in the para position an additional zone which is positive (P4) in the full antagonists and negative (M4) in the antagonists displaying intrinsic sympathomimetic activity (ISA). The MEP-based pharmacophoric models of PEAs, AOPAs, and oxime ethers show common elements and lead to a proposed general model for beta-adrenoceptor ligands.

Adrenergic beta-Agonists↗
PubMed2003

Differences in the compressive stress-strain response of infiltrating ductal carcinomas with and without lobular features--implications for mammography and elastography.

In a previous study, it was noted that in some cases when compressive strains greater than about 5% were applied to tumors removed from the breast, there was an abrupt and irreversible change in the tissue stiffness. The data from that study were further analyzed and infiltrating ductal carcinomas with and without lobular features were selected for additional testing to explore their behavior under compressive strains from 0-10%. Fresh tumor samples were tested using a servo-hydraulic Instron testing machine to apply ramp type displacement loads to the samples. The results show that when strains greater than 5% are applied to the tumor tissue without lobular features, there is an irreversible decrease in the stiffness of the tissue while no such change is noted in the other tumor tissue. The implications for this behavior in making mammographic and elastographic images of the breast were then explored using finite element simulations to determine under what compression conditions could the critical strain threshold be reached in the tumor tissue.

Breast Neoplasms↗
PubMed1998

Determination of rotations in three dimensions using two-dimensional portal image registration.

The relative relationships among anatomic features visualized on planar radiographic images change due to rotations of the patient out of the imaging plane. These changes can be predicted a priori from a three-dimensional radiographic model of the patient. In this study we assess the feasibility of using that information together with a planar image feature alignment tool to account for out-of-plane rotations in the evaluation of subsequent clinical patient images. A series of digitally reconstructed radiographs (DRRs) with known patient rotations was generated from a computed tomography scan of an anthropomorphic head phantom. Fixed anatomic features were extracted, as seen in the DRRs of rotated anatomy and entered into a database. Alignment of features from test radiographs with those from an entry in this database yielded an estimate of rotation out of plane (database entry that resulted in the best fit via planar transformation) along with the planar components of setup errors in the rotated plane. Tests using DRRs and films show that it is possible to select anatomic features in AP skull radiographs with position and orientation sensitive to out-of-plane rotation.

Databases as Topic↗
PubMed1998

Thrombolysis-related intracranial hemorrhage: a radiographic analysis of 244 cases from the GUSTO-1 trial with clinical correlation. Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries.

BACKGROUND AND PURPOSE: Intracranial hemorrhage (ICH) is a serious complication of thrombolytic therapy. We systematically reviewed the radiographic features of 244 cases of symptomatic ICH complicating thrombolysis for acute myocardial infarction in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO-1) trial, correlated these observations with clinical data, and speculated on hemorrhage pathogenesis. METHODS: CT scans from 244 patients suffering symptomatic ICH were systematically reviewed for selected radiographic features, including ICH type, location, hematoma characteristics, mass effect features, hydrocephalus, and preexisting lesions. Hematoma volume was estimated by computer-assisted volumetric analysis. Data from this analysis were correlated with clinical data including hypertension, anticoagulation, age, thrombolytic regimen, and ICH timing. RESULTS: Most hemorrhages were large (median [25th, 75th percentile] volume, 72 mL [39, 118]), solitary (66%), lobar (77%), confluent (80%), and intraparenchymal (82%) with a blood/fluid level (82%) and little edema (median [25th, 75th percentile] volume, 9 mL [5, 16]). Hydrocephalus (P<.001), any one mass effect feature (P<.001), intraventricular hemorrhage (P=.022), mottled hematoma appearance (P=.050), and hematoma blood/fluid level (P<.001) were associated with higher hemorrhage volume in the radiographic analysis, as were older age (P=.005), treatment with combined streptokinase and tissue plasminogen activator (P=.034), and hemorrhage onset 8 to 13 hours after treatment (P=.008) in the clinical analysis. Subdural hemorrhage was a high-volume subgroup whose risk increased with antecedent trauma (P=.026) or syncope (P=.006). Deep intraparenchymal hemorrhage was associated with hypertension (P=.016), and multifocal ICH occurred significantly earlier after treatment (P=.002). CONCLUSIONS: Although the majority of postthrombolytic ICH are large, solitary, and supratentorial, the spectrum is diverse. Features of mass effect reflected the large volumes, and hematoma characteristics of mottling and blood/fluid levels were frequent. Thrombolysis-related coagulopathy and age appear to be the most important identifiable factors in the genesis of postthrombolytic ICH, but the hemorrhage subtype seen may reflect an interaction with other factors such as hypertension, ICH timing, antecedent head trauma, and syncope.

Cerebral Hemorrhage↗
PubMed2026

Diffusion MRI radiomics in meningiomas: imaging correlates of tumor grade and intraoperative consistency.

OBJECTIVE: Despite advancements in imaging studies, the preoperative prediction of the biological behavior and intraoperative consistency of intracranial meningiomas remains limited. This study evaluated the association of volumetric diffusion-based and texture-derived radiomic features extracted from routine MRI with histopathological aggressiveness and intraoperative tumor consistency. METHODS: Ninety-seven intracranial meningiomas resected at two tertiary centers were retrospectively analyzed. Volumetric segmentation was performed on contrast-enhanced T1-weighted MRI and coregistered to apparent diffusion coefficient (ADC) maps. Data on first-order diffusion metrics and selected texture features were collected. The associations between World Health Organization (WHO) grade and Ki-67 index were assessed using nonparametric tests and Spearman correlation analysis. Independent factors associated with intraoperative tumor consistency (Zada grades 1-5) were evaluated via multivariate ordinal logistic regression analysis that adjusted for tumor volume, skull base location, calcification status, and WHO grade. Secondary receiver operating characteristic (ROC) curve analyses were performed to differentiate solid (Zada grades 4-5) from soft (Zada grades 1-2) tumors. ROC analyses were performed within the study cohort and were intended as exploratory assessments of discriminative performance. RESULTS: The mean ADC (ADCmean) and the 10th percentile of the ADC decreased significantly with increasing WHO grade (p < 0.001). ADCmean had a moderate inverse correlation with the Ki-67 index (r = -0.42, p < 0.001) and intraoperative tumor consistency (r = -0.45, p < 0.001). In the multivariate analysis, the ADCmean remained independently associated with increasing tumor firmness. Each 0.1 &#xd7; 10-3 mm2/sec increase corresponded to a 38% reduction in the odds of belonging to a higher consistency category (OR 0.62, 95% CI 0.51-0.74, p < 0.001). The ROC analysis showed good discrimination for solid tumors (area under the curve 0.847, 95% CI 0.742-0.953) and soft tumors (area under the curve 0.824, 95% CI 0.714-0.935). Texture features had weaker associations with intraoperative tumor consistency. CONCLUSIONS: Volumetric diffusion-derived metrics, particularly ADCmean, are associated with both histopathological aggressiveness and intraoperative tumor firmness in meningiomas. Diffusion imaging may reflect a graded microstructural continuum rather than a purely dichotomous property, providing complementary preoperative insights into surgical complexity.

Humans↗
PubMed1998

Diagnosing breast cancer from FNAs: variable relevance in neural network and logistic regression models.

We compared the selection of variables for building a classification model for the diagnosis of breast cancer using neural networks and logistic regression. A set of 460 cases was used to build neural network and logistic regression models that classify cell samples obtained by fine-needle aspiration (FNA) as malignant or benign, depending on nine pathology features. Variables selected by a step down logistic regression model were compared to those selected by a measure of relevance derived from neural network weights. Since both types of models resulted in similar predictive accuracy, we expected approximately the same variables to be selected. The variables with the highest relevance values for the neural network models corresponded to those of high significance in univariate logistic regression models, but were not the ones selected in the step down procedure of multivariate models. Variable relevance based on weights for neural network models does not seem to be a consistent index of the importance of that variable for multivariate models such as logistic regression.

Analysis of Variance↗
PubMed2003

Discrimination of benign common nevi from malignant melanoma lesions by use of features based on spectral properties of the wavelet transform.

OBJECTIVE: To evaluate the possibilities of describing and discriminating common nevi and malignant melanoma tissue with features based on spectral properties of the Daubechies 4 wavelet transform. STUDY DESIGN: Images of common nevi and malignant melanoma were dissected in square elements. The wavelet coefficients were calculated inside the square elements. The diagonal coefficients and related power spectra were used for further analysis. The analysis results served as guide for the selection of features, including standard deviations of wavelet coefficients inside the frequency bands and the energy of the frequency bands. These features describe properties of the frequency bands, representing information on different scales. To test the usefulness of the features for discrimination, a study set of 80 cases was classified by classification and regression trees analysis. The set was divided into a training set and a test set. RESULTS: In the case of benign common nevi, the energies of the lower frequency bands and higher, whereas malignant melanoma tissue shows more variability of the coefficients in higher-frequency bands. The influence on the detail properties of the images was studied by suppression of coefficients with low values, which are concentrated mainly in higher-frequency bands. In the case of benign common nevi the main information is contained in 15% of the coefficients and in the case of malignant melanoma, in 39%. The results of classification show a clear-cut difference between the cases. The classification correctly classified 95.78% of nevi elements and 94.22% of melanoma elements in the training set and 100% of cases of benign nevi and 80% of cases of malignant melanoma in the test set. CONCLUSION: Features based on the wavelet power spectrum contain sufficient information for differentiation between common nevi and malignant melanomas.

Artificial Intelligence↗
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↗
PubMed2005

Long term evaluation of disease progression through the quantitative magnetic resonance imaging of symptomatic knee osteoarthritis patients: correlation with clinical symptoms and radiographic changes.

The objective of this study was to further explore the cartilage volume changes in knee osteoarthritis (OA) over time using quantitative magnetic resonance imaging (qMRI). These were correlated with demographic, clinical, and radiological data to better identify the disease risk features. We selected 107 patients from a large trial (n = 1,232) evaluating the effect of a bisphosphonate on OA knees. The MRI acquisitions of the knee were done at baseline, 12, and 24 months. Cartilage volume from the global, medial, and lateral compartments was quantified. The changes were contrasted with clinical data and other MRI anatomical features. Knee OA cartilage volume losses were statistically significant compared to baseline values: -3.7 +/- 3.0% for global cartilage and -5.5 +/- 4.3% for the medial compartment at 12 months, and -5.7 +/- 4.4% and -8.3 +/- 6.5%, respectively, at 24 months. Three different populations were identified according to cartilage volume loss: fast (n = 11; -13.2%), intermediate (n = 48; -7.2%), and slow (n = 48; -2.3%) progressors. The predictors of fast progressors were the presence of severe meniscal extrusion (p = 0.001), severe medial tear (p = 0.005), medial and/or lateral bone edema (p = 0.03), high body mass index (p < 0.05, fast versus slow), weight (p < 0.05, fast versus slow) and age (p < 0.05 fast versus slow). The loss of cartilage volume was also slightly associated with less knee pain. No association was found with other Western Ontario McMaster Osteoarthritis Index (WOMAC) scores, joint space width, or urine biomarker levels. Meniscal damage and bone edema are closely associated with more cartilage volume loss. These data confirm the significant advantage of qMRI for reliably measuring knee structural changes at as early as 12 months, and for identifying risk factors associated with OA progression.

Aged↗
PubMed1993

Sexual selection, Machiavellian intelligence, and the origins of psychosis.

According to Darwin's theory of sexual selection some features that differentiate the two sexes evolve by a process of "male competition" and "female choice". The sex difference in age of onset of psychotic illness in man may relate to a sexual dimorphism in cerebral organisation (the male brain being more lateralised or asymmetrical than the female brain), a difference consistent with a role for sexual selection in the evolution of the human brain. Differing criteria (reflected in a cross-culturally stable difference in mean age at marriage) in males and females for selecting personality characteristics in a mate may generate diversity in the balance of growth between the hemispheres, and this could maintain the high and relatively constant rates of psychosis in human populations.

Brain↗
PubMed1995

Structure-activity and structure-selectivity studies on diaminoquinazolines and other inhibitors of Pneumocystis carinii and Toxoplasma gondii dihydrofolate reductase.

Twenty-eight 2,4-diaminopteridines with alkyl and aralkyl groups at the 6- and 7-positions, five 1,3-diamino-7,8,9,10-tetrahydropyrimido [4,5-c]isoquinolines with an alkyl, alkylthio, or aryl group at the 6-position, and nine 4,6-diamino-1,2-dihydro-s-triazines with one or two alkyl groups at the 2-position and a substituted phenyl or naphthyl group at the 1-position were evaluated as inhibitors of dihydrofolate reductase enzymes from Pneumocystis carinii, Toxoplasma gondii, and rat liver. Halogen substitution at the 5- or 6-position of 2,4-diaminoquinazoline favored selective binding to the P. carinii enzyme but not the T. gondii enzyme. For example, the 50% inhibitory concentrations of 2,4-diamino-6-chloroquinazoline as an inhibitor of P. carinii, T. gondii, and rat liver dihydrofolate reductase were 3.6, 14 and 29 microM, respectively, corresponding to 12-fold selectivity for the P. carinii enzyme but only marginal selectivity for the T. gondii enzyme. Greater than fivefold selectivity for P. carinii but not T. gondii dihydrofolate reductase was also observed for the 2,4-diaminoquinazolines with 5-methyl, 5-fluoro, 5- and 6-bromo, 6-chloro, and 5-chloro-6-bromo substitution. In contrast, alkyl and aralkyl substitution at the 6- and 7-positions of 2,4-diaminopteridines was found to be a favorable feature for selective inhibition of the T. gondii enzyme and, in two cases, for both enzymes. Nine of the fifty-one compounds tested against P. carinii dihydrofolate reductase and four of the thirty compounds tested against T. gondii dihydrofolate reductase displayed fivefold or greater selectivity for the microbial enzyme versus the rat liver enzyme. The most selective against both enzymes was 2,4-diamino-6,7-bis(cyclohexylmethyl) pteridine, with a selectivity ratio 2 orders of magnitude greater than the value reported for trimetrexate and piritrexim. Since substitution at the 7-position is generally considered to be detrimental to the binding of 2,4-diaminop-teridines and related compounds to mammalian dihydrofolate reductase, the selectivity observed in this study with the 6,7-bis(cyclohexylmethyl) analog may represent a useful approach to enhancing selective inhibition of the enzyme from nonmammalian species.

Animals↗