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 217 records · Page 12Linked to original sources
PubMed2000

Serum C-peptide concentrations poorly phenotype type 2 diabetic end-stage renal disease patients.

BACKGROUND: A homogeneous patient population is necessary to identify genetic factors that regulate complex disease pathogenesis. In this study, we evaluated clinical and biochemical phenotyping criteria for type 2 diabetes in end-stage renal disease (ESRD) probands of families in which nephropathy is clustered. C-peptide concentrations accurately discriminate type 1 from type 2 diabetic patients with normal renal function, but have not been extensively evaluated in ESRD patients. We hypothesized that C-peptide concentrations may not accurately reflect insulin synthesis in ESRD subjects, since the kidney is the major site of C-peptide catabolism and would poorly correlate with accepted clinical criteria used to classify diabetics as types 1 and 2. METHODS: Consenting diabetic ESRD patients (N = 341) from northeastern Ohio were enrolled. Clinical history was obtained by questionnaire, and predialysis blood samples were collected for C-peptide levels from subjects with at least one living diabetic sibling (N = 127, 48% males, 59% African Americans). RESULTS: Using clinical criteria, 79% of the study population were categorized as type 1 (10%) or type 2 diabetics (69%), while 21% of diabetic ESRD patients could not be classified. In contrast, 98% of the patients were classified as type 2 diabetics when stratified by C-peptide concentrations using criteria derived from the Diabetes Control and Complications Trial Research Group (DCCT) and UREMIDIAB studies. Categorization was concordant in only 70% of ESRD probands when C-peptide concentration and clinical classification algorithms were compared. Using clinical phenotyping criteria as the standard for comparison, C-peptide concentrations classified diabetic ESRD patients with 100% sensitivity, but only 5% specificity. The mean C-peptide concentrations were similar in diabetic ESRD patients (3.2 +/- 1.9 nmol/L) and nondiabetic ESRD subjects (3.5 +/- 1.7 nmol/L, N = 30, P = NS), but were 2.5-fold higher compared with diabetic siblings (1.3 +/- 0.7 nmol/L, N = 30, P < 0.05) with normal renal function and were indistinguishable between type 1 and type 2 diabetics. Although 10% of the diabetic ESRD study population was classified as type 1 diabetics using clinical criteria, only 1.5% of these patients had C-peptide levels less than 0.20 nmol/L, the standard cut-off used to discriminate type 1 from type 2 diabetes in patients with normal renal function. However, the criteria of C-peptide concentrations> 0.50 nmol/L and diabetes onset in patients who are more than 38 years old identify type 2 diabetes with a 97% positive predictive value in our ESRD population. CONCLUSIONS: Accepted clinical criteria, used to discriminate type 1 and type 2 diabetes, failed to classify a significant proportion of diabetic ESRD patients. In contrast to previous reports, C-peptide levels were elevated in the majority of type 1 ESRD diabetic patients and did not improve the power of clinical parameters to separate them from type 2 diabetic or nondiabetic ESRD subjects. Accurate classification of diabetic ESRD patients for genetic epidemiological studies requires both clinical and biochemical criteria, which may differ from norms used in diabetic populations with normal renal function.

Adult↗
PubMed2004

Development of biomarkers based on diet-dependent metabolic serotypes: characteristics of component-based models of metabolic serotypes.

Our research seeks to identify a serum profile, or serotype, that reflects the systemic physiologic modifications resultant from dietary restriction (DR), in part such that this knowledge can be applied for biomarker studies. Direct comparison suggests that component-based classification algorithms consistently out-perform distance-based metrics for studies of nutritional modulation of metabolic serotype, but are subject to over-fitting concerns. Intercohort differences in the sera metabolome could partially obscure the effects of DR. Further analysis now shows that implementation of component-based approaches (also called projection methods) optimized for class separation and controlled for over-fitting have >97% accuracy for distinguishing sera from control or DR rats. DR's effect on the metabolome is shown to be robust across cohorts, but differs in males and females (although some metabolites are affected in both). We demonstrate the utility of projection-based methods for both sample and variable diagnostics, including identification of critical metabolites and samples that are atypical with respect to both class and variable models. Inclusion of non-statistically different variables enhances classification models. Variables that contribute to these models are sharply dependent on mathematical processing techniques; some variables that do not contribute under one paradigm are powerful under alternative mathematical paradigms. In practical terms, this information may find purpose in other endeavors, such as mechanistic studies of DR. Application of these approaches confirms the utility of megavariate data analysis techniques for optimal generation of biomarkers based on nutritional modulation of physiological processes.

Animals↗
PubMed2004

Fingerprint classification using a feedback-based line detector.

We present a fingerprint classification algorithm in this paper. This algorithm classifies a fingerprint image into one of the five classes: Arch, Left loop, Right loop, Whorl, and Tented arch. We use a new low-dimensional feature vector obtained from the output of a novel oriented line detector presented here. Our line detector is a co-operative dynamical system that gives oriented lines and preserves multiple orientations at points where differently oriented lines meet. Our feature extraction process is based on characterizing the distribution of orientations around the fingerprint. We discuss three different classifiers: support vector machines, nearest-neighbor classifier, and neural network classifier. We present results obtained on a National Institute of Standards and Technology (NIST) fingerprint database and compare with other published results on NIST databases. All our classifiers perform equally well, and this suggests that our novel line detection and feature extraction process indeed captures all the crucial information needed for classification in this problem.

Journal Article↗
PubMed2005

Texture signals in whisker vibrations.

Rodents excel in making texture judgments by sweeping their whiskers across a surface. Here we aimed to identify the signals present in whisker vibrations that give rise to such fine sensory discriminations. First, we used sensors to capture vibration signals in metal whiskers during active whisking of an artificial system and in natural whiskers during whisking of rats in vivo. Then we developed a classification algorithm that successfully matched the vibration frequency spectra of single trials to the texture that induced it. For artificial whiskers, the algorithm correctly identified one texture of eight alternatives on 40% of trials; for in vivo natural whiskers, the algorithm correctly identified one texture of five alternatives on 80% of trials. Finally, we asked which were the key discriminative features of the vibration spectra. Under both artificial and natural conditions, the combination of two features accounted for most of the information: The modulation power-the power of the part of the whisker movement representing the modulation due to the texture surface-increased with the coarseness of the texture; the modulation centroid-a measure related to the center of gravity within the power spectrum-decreased with the coarseness of the texture. Indeed, restricting the signal to these two parameters led to performance three-fourths as high as the full spectra. Because earlier work showed that modulation power and centroid are directly related to neuronal responses in the whisker pathway, we conclude that the biological system optimally extracts vibration features to permit texture classification.

Action Potentials↗
PubMed2006

Trainable fusion rules. I. Large sample size case.

A wide selection of standard statistical pattern classification algorithms can be applied as trainable fusion rules while designing neural network ensembles. A focus of the present two-part paper is finite sample effects: the complexity of base classifiers and fusion rules; the type of outputs provided by experts to the fusion rule; non-linearity of the fusion rule; degradation of experts and the fusion rule due to the lack of information in the design set; the adaptation of base classifiers to training set size, etc. In the first part of this paper, we consider arguments for utilizing continuous outputs of base classifiers versus categorical outputs and conclude: if one succeeds in having a small number of expert networks working perfectly in different parts of the input feature space, then crisp outputs may be preferable over continuous outputs. Afterwards, we oppose fixed fusion rules versus trainable ones and demonstrate situations where weighted average fusion can outperform simple average fusion. We present a review of statistical classification rules, paying special attention to these linear and non-linear rules, which are employed rarely but, according to our opinion, could be useful in neural network ensembles. We consider ideal and sample-based oracle decision rules and illustrate characteristic features of diverse fusion rules by considering an artificial two-dimensional (2D) example where the base classifiers perform well in different regions of input feature space.

Animals↗
PubMed2006

Blunt injury to the thyroid gland: proposed classification and treatment algorithm.

BACKGROUND: Traumatic hemorrhage of the thyroid gland is a rare injury with few published case reports. Surgical and nonsurgical management have been advocated but there is no consensus of opinion. METHODS: This study is a retrospective review of published case reports including two own cases. A classification and an algorithm for diagnosis and treatment of thyroid gland injuries is proposed. RESULTS: A literature review reveals 16 case reports. Of these, 11 patients underwent neck exploration for control of hemorrhage and resection of the associated thyroid abnormality while 5 patients were conservatively managed. Of note is the fact that 10 of the 16 patients (63%) had pre-existing thyroid disease. These 10 patients constituted (91%) of the group requiring surgical treatment. CONCLUSION: Our proposed classification and algorithm offers management guidelines for this rather rare injury. Conservative treatment may be successful in selected patients with lower grade injuries and without concomitant thyroid disease.

Algorithms↗
PubMed2003

Multidimensional subgroups in migraine: differential treatment outcome to a pain medicine program.

OBJECTIVE: The present study compared two different approaches for deriving patient profiles on their ability to predict treatment outcome to a pain medicine program for migraine headache. DESIGN/METHODS: Using visual analog scale measures of pain intensity and functional limitations and the Beck Depression Inventory (BDI), as a measure of depression, 235 migraine patients were classified into statistical clusters. The same patients were also classified using the Multidimensional Pain Inventory (MPI) algorithm into three subgroups: Adaptive copers (AC), characterized by lower reported levels of pain intensity, life interference, and distress, as well as higher levels of perceived life control; interpersonally distressed (ID), characterized by more intermediate levels of pain, distress, and interference, with a predominant perception of inadequate support and punishing responses from significant others; and dysfunctional (Dys), characterized by high levels of pain severity, life interference, and distress and low levels of perceived life control and activity. RESULTS: The results of the K-cluster analysis yielded a three-cluster solution: The low impact cluster, was characterized by low pain, low functional limitations and low depression and showed significant reductions in pre-to-posttreatment pain; the moderate impact cluster displayed higher levels of pain and functional limitations and low depression and showed only slight pre-to-posttreatment pain reduction; and the high impact cluster displayed the highest levels of pain, functional limitations, and depression and showed significant increases in pre-to-posttreatment pain. Unlike the K-clustered groups, MPI subgroups failed to differentially predict treatment outcome. When the K-clustered groups were crosstabulated with the MPI subgroups, the predictive validity of the MPI subgroups was enhanced. CONCLUSION: This study questions the validity of the MPI subgroup classification algorithm. The results indicate that the K-clustering approach is more useful than the MPI in deriving meaningful patient clusters that differentially predict treatment outcome in a migraine population.

Adolescent↗
PubMed2026

sWGS Identifies a Copy-Number-High Subset of TP53-mutated Multiple-Classifier Endometrial Carcinomas With Adverse Clinicopathological Features.

TP53-mutated "multiple-classifier" endometrial carcinomas represent a diagnostically challenging subgroup within current molecular classification algorithms. Although these tumors are assigned to POLE-mutated or mismatch repair-deficient categories according to current ESGO/FIGO-based algorithms, their biological heterogeneity remains incompletely characterized. Herein, we retrospectively analyzed TP53-mutated multiple-classifier endometrial carcinomas identified through routine molecular profiling at our institution between 2022 and 2025 using an integrated histopathological, immunohistochemical, targeted sequencing, and shallow whole-genome sequencing approach. Copy-number alteration-high (CNA-high) status was defined as &#x2265;5 large-scale genomic alterations, corresponding to copy-number gains or losses &#x2265;3 Mb within a single chromosomal arm excluding whole-arm alterations. Among 33 analyzable TP53-mutated multiple-classifier endometrial carcinomas, sWGS identified 12 CNA-high tumors (36.4%) and 21 CNA-low tumors (63.6%). CNA-high tumors were more frequently non-endometrioid, high-grade, and advanced-stage according to FIGO 2023. They showed higher TP53 variant allele frequencies (VAF) and higher TP53 VAF-to-tumor-cellularity ratios. After a median follow-up of 12.8 months, recurrences (6/33; 18.2%) and disease-related deaths (3/33; 9.1%) were observed in the CNA-high subgroup, whereas no recurrence or disease-related death was observed among CNA-low patients. These findings indicate that TP53-mutated multiple-classifier endometrial carcinomas comprise biologically distinct subsets that are not fully captured by current 4-tier TCGA-based molecular classification and ESGO-based risk stratification. In this cohort, sWGS identified a CNA-high group with adverse clinicopathological features and clinical events suggesting a potentially more aggressive clinical course. Integration of genome-wide copy-number profiling may therefore refine the biological interpretation of TP53 alterations in multiple-classifier endometrial carcinomas and warrants validation in larger multicenter cohorts.

TP53↗
PubMed1999

Neurocognitive deficits in medulloblastoma survivors and white matter loss.

Although previous studies have documented a significant risk of intellectual loss after treatment for childhood medulloblastoma (MED), the pathophysiology underlying this process is poorly understood. The purpose of this study was to test the hypotheses that (1) patients treated for MED in childhood have reduced volumes of normal white matter (NWM) related to their treatment with craniospinal irradiation with or without chemotherapy, and (2) deficits in NWM among patients surviving MED can at least partially explain deficits in their intellectual performance. Eighteen pediatric patients previously treated for MED were matched on the basis of age at the time of evaluation to 18 patients previously treated for low-grade posterior fossa tumors with surgery alone (mean difference, 3.7 months). Evaluations were conducted with age-appropriate neurocognitive testing and quantitative magnetic resonance imaging by using a novel automated segmentation and classification algorithm constructed from a hybrid neural network. Patients treated for MED had significantly less NWM (p < 0.01) and significantly lower Full-Scale IQ values than those treated for low-grade tumors (mean, 82.1 vs 92.9). In addition, NWM had a positive and statistically significant association with Full-Scale IQ among the patients treated for MED. We conclude that irradiation- or chemotherapy-induced destruction of NWM can at least partially explain intellectual and academic achievement deficits among MED survivors.

Adolescent↗
PubMed1995

Variability in absolute lymphocyte counts obtained by automated cell counters.

There is increasing interest in the absolute lymphocyte count. This is partly driven by the need to obtain absolute values for lymphocyte subsets such as absolute CD4+ counts in human immunodeficiency virus (HIV)-infected persons. The absolute total lymphocyte count is usually determined in the routine hematology laboratory on a separate sample from the same patient specimen and then combined with percentage results from flow cytometry to obtain the absolute value of the lymphocyte subsets. We have studied analytic variability in the absolute lymphocyte determination and compared it to the variability of the total white blood count (WBC). In a series of 524 specimens, four different automated methods were compared to each other and to the traditional eye count differential. The automated methods were four widely used automated cell counters (Technicon H*1, TOA NE8000, Coulter STKS, and Abbott CD3000). The results indicate that analytic variability in the absolute lymphocyte counts, due, primarily, to method variability, is significant and is larger than the variability typically observed on interlaboratory trials of relative CD4 counts. These method biases cannot easily be reduced by calibration, since the cell classification algorithms are built-in features of the various cell counters. Analytic variability of the absolute lymphocyte counts was found to be 12.4% compared with analytic variability of only 4.9% for total WBC counts on the same samples. Our data suggest that more precise results would be obtained if flow cytometry results expressed each phenotype as a fraction of the leukocytes as well as total lymphocytes. Conversion to absolute values could then be accomplished through determination of the total WBC in the routine hematology laboratory.

Automation↗
PubMed1991

Design and evaluation of a fiberoptic fluorescence guided laser recanalization system.

Current angioplasty techniques for recanalization of totally occluded arteries are limited by the inability to cross the occlusion and by the risk of perforation. A fiberoptic fluorescence guided laser recanalization system was developed and evaluated in vitro for recanalization of 17 human femoral or tibial totally occluded arterial segments (length 1.9-6.8 cm, diameter 2.5-6.0 mm). A 400 or 600 micron silica fiber was coupled to a helium-cadmium laser (lambda = 325 nm) for fluorescence excitation and to a holmium: YAG laser (lambda = 2.1 micron) for tissue ablation. Fluorescence was recorded during recanalization after every other holmium laser pulse. During recanalization, each arterial segment was bent 30-90 degrees with respect to the fiber to simulate arterial tortuosity. Ablation continued with fiber advancement as long as the fluorescence confirmed that the target tissue was atherosclerotic. Arterial spectra were classified as normal or atherosclerotic by an on-line computerized fluorescence classification algorithm (sensitivity 93%, specificity 95%). Normal fluorescence necessitated redirection of the fiber greater than 30 times per segment to continue recanalization. Fifteen of 17 totally occluded arteries had multiple recanalization channels created following total energy delivery of 40-1,016 Joules per segment with no angiographic or histologic evidence of laser perforation. Two heavily calcified arterial occlusions were not recanalized due to inhibition of holmium: YAG laser ablation by the recording of normal fluorescence spectra. Therefore, this fluorescence guided laser recanalization system appears safe and effective for recanalization of totally occluded arteries and merits in vivo evaluation. However, the lower sensitivity of fluorescence detection of heavily calcified plaques may limit the efficacy (but not safety) of fluorescence guided recanalization of heavily calcified occlusions.

Angioplasty, Laser↗
PubMed2006

Will my protein crystallize? A sequence-based predictor.

We propose a machine-learning approach to sequence-based prediction of protein crystallizability in which we exploit subtle differences between proteins whose structures were solved by X-ray analysis [or by both X-ray and nuclear magnetic resonance (NMR) spectroscopy] and those proteins whose structures were solved by NMR spectroscopy alone. Because the NMR technique is usually applied on relatively small proteins, sequence length distributions of the X-ray and NMR datasets were adjusted to avoid predictions biased by protein size. As feature space for classification, we used frequencies of mono-, di-, and tripeptides represented by the original 20-letter amino acid alphabet as well as by several reduced alphabets in which amino acids were grouped by their physicochemical and structural properties. The classification algorithm was constructed as a two-layered structure in which the output of primary support vector machine classifiers operating on peptide frequencies was combined by a second-level Naive Bayes classifier. Due to the application of metamethods for cost sensitivity, our method is able to handle real datasets with unbalanced class representation. An overall prediction accuracy of 67% [65% on the positive (crystallizable) and 69% on the negative (noncrystallizable) class] was achieved in a 10-fold cross-validation experiment, indicating that the proposed algorithm may be a valuable tool for more efficient target selection in structural genomics. A Web server for protein crystallizability prediction called SECRET is available at http://webclu.bio.wzw.tum.de:8080/secret.

Amino Acid Sequence↗
PubMed1994

DNA sequence confidence estimation.

A significant bottleneck in the current DNA sequencing process is the manual editing of trace data generated by automated DNA sequencers. This step is used to correct base calls and to associate to each base call a confidence level. The confidence levels are used in the assembly process to determine overlaps and to resolve discrepancies in determining the consensus sequence. This single step may cost as much as 4 to 8 cents per finished base. We report an approach to automated trace editing using classification trees to detect and exploit context-based patterns in trace peak heights. Local base composition and nearby peak heights account for 80% of the variations in peak heights. Classification algorithms were developed to identify 37% of automated base calls that differ from the consensus sequence. With these algorithms, 12% of the base calls had confidence levels less than 90%.

Algorithms↗
PubMed1995

Spatiotemporal stability and patterning of speech movement sequences.

In order to examine the stability and patterning of speech movement sequences, movements of the lip were recorded as subjects produced a phrase at normal, fast, and slow rates. Three methods of analysis were employed. First, a new index of spatiotemporal stability was derived by summing the standard deviations computed across amplitude- and time-normalized displacement records. This index indicated that normal and fast rates of speech production result in more stable movement execution compared to slow rates. In the second analysis, the relative time of occurrence of the peak velocity of the three middle opening movements of the utterance was measured. For each of the three peaks, the preservation of relative timing was assessed by applying Genter's (1987) slope test. The results clearly indicate that the relative timing of these events does not remain constant across changes in speech rate. The relative timing of the middle opening gestures shifted, becoming later as utterance duration increased. In a third analysis, pattern recognition techniques were applied to the normalized displacement waveforms. A classification algorithm was highly successful in sorting waveforms into normal, fast, and slow rate conditions. These findings were interpreted to suggest that, within a subject, three distinct patterns or movement templates exist, one for each rate of production. Speech rate appears to be a global parameter, one that affects the entire command sequence for the utterance.

Adult↗
PubMed1997

Detection of life-threatening cardiac arrhythmias using the wavelet transformation.

Time-frequency wavelet theory is used for the detection of life threatening electrocardiography (ECG) arrhythmias. This is achieved through the use of the raised cosine wavelet transform (RCWT). The RCWT is found to be useful in differentiating between ventricular fibrillation, ventricular tachycardia and atrial fibrillation. Ventricular fibrillation is characterised by continuous bands in the range of 2-10 Hz; ventricular tachycardia is characterised by two distinct bands: the first band in the range of 2-5 Hz and the second in the range of 6-8 Hz; and atrial fibrillation is determined by a low frequency band in the range of 0-5 Hz. A classification algorithm is developed to classify ECG records on the basis of the computation of three parameters defined in the time-frequency plane of the wavelet transform. Furthermore, the advantage of localising and separating ECG signals from high as well as intermediate frequencies is demonstrated. The above capabilities of the wavelet technique are supported by results obtained from ECG signals obtained from normal and abnormal subjects.

Algorithms↗
PubMed1993

Analytical methods to differentiate similar electroencephalographic spectra: neural network and discriminant analysis.

Differences in electroencephalographic (EEG) power spectra obtained under similar, but not identical, conditions may be difficult to discern using standard techniques. Statistical analysis may not be useful because of the large number of comparisons necessary. Visual recognition of differences also may be difficult. A new technique, neural network analysis, has been used successfully in other problems of pattern recognition and classification. We examined a number of methods of classifying similar EEG data: standard statistical analysis (analysis of variance), visual recognition, discriminant analysis, and neural network analysis. Twenty-nine volunteers received either thiopental (n = 9), midazolam (n = 10), or propofol (n = 10) in sedative doses in 3 different studies. These drugs produced very similar changes in the EEG power spectra. Except for beta 2 power during thiopental infusion, differences between drugs could not be detected using analysis of variance. Visual categorization was correct in 72% of the baseline EEGs, 70% of thiopental EEGs, 27% of propofol EEGs, and 46% of midazolam EEGs. A classification neural network (Learning Vector Quantization network) containing a Kohonen hidden layer was able to successfully classify 57 of 58 EEG samples (of 4 minutes' duration). Discriminant analysis had a similar rate of success. This level of performance was achieved by dividing the EEG power spectrum from 1 to 30 Hz into 15 2-Hz bandwidths. When the EEG power spectrum was divided into the "classical" frequency bandwidths (alpha, beta 1, beta 2, theta, delta), both neural network and discriminant analysis performance deteriorated. By training the network using only certain inputs we were able to identify drug-specific bandwidths that seemed to be important in correct classification. We conclude that propofol, thiopental, and midazolam produce different effects on the EEG and that both neural network and discriminant analysis are useful in identifying these differences. We also conclude that EEG spectra should be analyzed without using classical EEG bands (alpha, beta, etc.). Additionally, neural networks can be used to identify frequency bands that are "important" in specific drug effects on the EEG. Once a classification algorithm is obtained using either a neural network or discriminant analysis, it could be used as an on-line monitor to recognize drug-specific EEG patterns.

Adult↗
PubMed2006

Bank note recognition for the vision impaired.

Blind Australians find great difficulty in recognising bank notes. Each note has the same feel, with no Braille markings, irregular edges or other tangible features. In Australia, there is only one device available that can assist blind people recognise their notes. Internationally, there are devices available; however they are expensive, complex and have not been developed to cater for Australian currency. This paper discusses a new device, the MoneyTalker that takes advantage of the largely different colours and patterns on each Australian bank note and recognises the notes electronically, using the reflection and transmission properties of light. Different coloured lights are transmitted through the inserted note and the corresponding sensors detect distinct ranges of values depending on the colour of the note. Various classification algorithms were studied and the final algorithm was chosen based on accuracy and speed of recognition. The MoneyTalker has shown an accuracy of more than 99%. A blind subject has tested the device and believes that it is usable, compact and affordable. Based on the devices that are available currently in Australia, the MoneyTalker is an effective alternative in terms of accuracy and usability.

Algorithms↗
PubMed2004

Use of near infrared spectroscopy in a study of binding media used in paintings.

The present study illustrates the use of near infrared spectroscopy in the conservation and restoration field. This application of the technique, widely used in other fields such as the agricultural and food industries, is very interesting due to two of its features: rapidity and non-destructiveness. In this study, three classification algorithms have been used to attempt to correlate the spectra of the samples under study--mixtures of two pigments (one organic and the other inorganic)--with binder recipes found in literature. Some of these recipes have very similar compositions, but they were distinguished into groups, depending on the painting technique associated with them.

Adhesives↗