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 199 records · Page 11Linked to original sources
PubMed2005

Optimizing the kernel in the empirical feature space.

In this paper, we present a method of kernel optimization by maximizing a measure of class separability in the empirical feature space, an Euclidean space in which the training data are embedded in such a way that the geometrical structure of the data in the feature space is preserved. Employing a data-dependent kernel, we derive an effective kernel optimization algorithm that maximizes the class separability of the data in the empirical feature space. It is shown that there exists a close relationship between the class separability measure introduced here and the alignment measure defined recently by Cristianini. Extensive simulations are carried out which show that the optimized kernel is more adaptive to the input data, and leads to a substantial, sometimes significant, improvement in the performance of various data classification algorithms.

Empirical Research↗
PubMed2006

Diagnosis of breast cancer using diffuse reflectance spectroscopy: Comparison of a Monte Carlo versus partial least squares analysis based feature extraction technique.

BACKGROUND AND OBJECTIVE: We explored the use of diffuse reflectance spectroscopy in the ultraviolet-visible (UV-VIS) spectrum for the diagnosis of breast cancer. A physical model (Monte Carlo inverse model) and an empirical model (partial least squares analysis) based approach, were compared for extracting diagnostic features from the diffuse reflectance spectra. STUDY DESIGN/METHODS: The physical model and the empirical model were employed to extract features from diffuse reflectance spectra measured from freshly excised breast tissues. A subset of extracted features obtained using each method showed statistically significant differences between malignant and non-malignant breast tissues. These features were separately input to a support vector machine (SVM) algorithm to classify each tissue sample as malignant or non-malignant. RESULTS AND CONCLUSIONS: The features extracted from the Monte Carlo based analysis were hemoglobin saturation, total hemoglobin concentration, beta-carotene concentration and the mean (wavelength averaged) reduced scattering coefficient. Beta-carotene concentration was positively correlated and the mean reduced scattering coefficient was negatively correlated with percent adipose tissue content in normal breast tissues. In addition, there was a statistically significant decrease in the beta-carotene concentration and hemoglobin saturation, and a statistically significant increase in the mean reduced scattering coefficient in malignant tissues compared to non-malignant tissues. The features extracted from the partial least squares analysis were a set of principal components. A subset of principal components showed that the diffuse reflectance spectra of malignant breast tissues displayed an increased intensity over wavelength range of 440-510 nm and a decreased intensity over wavelength range of 510-600 nm, relative to that of non-malignant breast tissues. The diagnostic performance of the classification algorithms based on both feature extraction techniques yielded similar sensitivities and specificities of approximately 80% for discriminating between malignant and non-malignant breast tissues. While both methods yielded similar classification accuracies, the model based approach provided insight into the physiological and structural features that discriminate between malignant and non-malignant breast tissues.

Adipose Tissue↗
PubMed2005

A study on several machine-learning methods for classification of malignant and benign clustered microcalcifications.

In this paper, we investigate several state-of-the-art machine-learning methods for automated classification of clustered microcalcifications (MCs). The classifier is part of a computer-aided diagnosis (CADx) scheme that is aimed to assisting radiologists in making more accurate diagnoses of breast cancer on mammograms. The methods we considered were: support vector machine (SVM), kernel Fisher discriminant (KFD), relevance vector machine (RVM), and committee machines (ensemble averaging and AdaBoost), of which most have been developed recently in statistical learning theory. We formulated differentiation of malignant from benign MCs as a supervised learning problem, and applied these learning methods to develop the classification algorithm. As input, these methods used image features automatically extracted from clustered MCs. We tested these methods using a database of 697 clinical mammograms from 386 cases, which included a wide spectrum of difficult-to-classify cases. We analyzed the distribution of the cases in this database using the multidimensional scaling technique, which reveals that in the feature space the malignant cases are not trivially separable from the benign ones. We used receiver operating characteristic (ROC) analysis to evaluate and to compare classification performance by the different methods. In addition, we also investigated how to combine information from multiple-view mammograms of the same case so that the best decision can be made by a classifier. In our experiments, the kernel-based methods (i.e., SVM, KFD, and RVM) yielded the best performance (Az = 0.85, SVM), significantly outperforming a well-established, clinically-proven CADx approach that is based on neural network (Az = 0.80).

Algorithms↗
PubMed2005

Classifying noisy protein sequence data: a case study of immunoglobulin light chains.

SUMMARY: The classification of protein sequences obtained from patients with various immunoglobulin-related conformational diseases may provide insight into structural correlates of pathogenicity. However, clinical data are very sparse and, in the case of antibody-related proteins, the collected sequences have large variability with only a small subset of variations relevant to the protein pathogenicity (function). On this basis, these sequences represent a model system for development of strategies to recognize the small subset of function-determining variations among the much larger number of primary structure diversifications introduced during evolution. Under such conditions, most protein classification algorithms have limited accuracy. To address this problem, we propose a support vector machine (SVM)-based classifier that combines sequence and 3D structural averaging information. Each amino acid in the sequence is represented by a set of six physicochemical properties: hydrophobicity, hydrophilicity, volume, surface area, bulkiness and refractivity. Each position in the sequence is described by the properties of the amino acid at that position and the properties of its neighbors in 3D space or in the sequence. A structure template is selected to determine neighbors in 3D space and a window size is used to determine the neighbors in the sequence. The test data consist of 209 proteins of human antibody immunoglobulin light chains, each represented by aligned sequences of 120 amino acids. The methodology is applied to the classification of protein sequences collected from patients with and without amyloidosis, and indicates that the proposed modified classifiers are more robust to sequence variability than standard SVM classifiers, improving classification error between 5 and 25% and sensitivity between 9 and 17%. The classification results might also suggest possible mechanisms for the propensity of immunoglobulin light chains to amyloid formation.

Algorithms↗
PubMed1989

Development and evaluation of a survey-based type of benefit classification for the Social Security program.

This attempt to classify Social Security beneficiaries by type of benefit using the new Survey of Income and Program Participation (SIPP) has yielded promising results. Evaluations of the classification algorithm based on comparison of the estimated number of beneficiaries in each of the several categories to independent estimates of the number of beneficiaries indicate that in most instances a high percentage of each category has been identified. For the most part, age and monthly benefit amount size distributions seem reasonable. Furthermore, very few persons in the sample who were identified as Social Security beneficiaries could not be assigned to one or another of the benefit groups. The classification procedure also represents a marked improvement over earlier efforts to classify type of beneficiary that relied on data from the March Supplement to the Current Population Survey (CPS). Most importantly, the benefit classification scheme based on SIPP data appears to provide reasonably reliable distinctions between retired-worker and widow benefits for widowed women and permits the identification of retired-worker benefits for those women dually entitled to retired-worker and spouse benefits. In addition, the distinction between disabled- and retired-worker benefits for recipients aged 62-64 appears to be reasonably reliable, and for women under age 65, the classification procedure distinguishes between disabled-worker benefits on the one hand and widow and widowed mother benefits on the other. Finally, SIPP procedures for identifying minor child beneficiaries yield markedly better estimates than those available from the Current Population Survey. These improvements in the SIPP context are due entirely to the presence of information not collected in the CPS. The enhancement of the SIPP data set in turn resulted directly from an assessment of earlier work carried out by Projector and Bretz in the CPS context and on extensive research into the nature of Social Security reporting errors in the CPS. The superiority of the SIPP data set is linked principally to the presence of three pieces of information: the Medicare BIC, the direct question on reasons for benefit receipt asked of persons under age 65, and the direct measurement of recipiency and amount of benefits for minor children. Other items of some import include self-reported work disability, retirement status (ever retired from a job), previous marital status for currently married women, age first prevented from working due to a health condition, and Supplemental Security Income misreporting items.(ABSTRACT TRUNCATED AT 400 WORDS)

Adolescent↗
PubMed2003

Reducing false positives in molecular pattern recognition.

In the search for new cancer subtypes by gene expression profiling, it is essential to avoid misclassifying samples of unknown subtypes as known ones. In this paper, we evaluated the false positive error rates of several classification algorithms through a 'null test' by presenting classifiers a large collection of independent samples that do not belong to any of the tumor types in the training dataset. The benchmark dataset is available at www2.genome.rcast.u-tokyo.ac.jp/pm/. We found that k-nearest neighbor (KNN) and support vector machine (SVM) have very high false positive error rates when fewer genes (<100) are used in prediction. The error rate can be partially reduced by including more genes. On the other hand, prototype matching (PM) method has a much lower false positive error rate. Such robustness can be achieved without loss of sensitivity by introducing suitable measures of prediction confidence. We also proposed a cluster-and-select technique to select genes for classification. The nonparametric Kruskal-Wallis H test is employed to select genes differentially expressed in multiple tumor types. To reduce the redundancy, we then divided these genes into clusters with similar expression patterns and selected a given number of genes from each cluster. The reliability of the new algorithm is tested on three public datasets.

Computational Biology↗
PubMed2001

Rapid prototyping of an EEG-based brain-computer interface (BCI).

The electroencephalogram (EEG) is modified by motor imagery and can be used by patients with severe motor impairments (e.g., late stage of amyotrophic lateral sclerosis) to communicate with their environment. Such a direct connection between the brain and the computer is known as an EEG-based brain-computer interface (BCI). This paper describes a new type of BCI system that uses rapid prototyping to enable a fast transition of various types of parameter estimation and classification algorithms to real-time implementation and testing. Rapid prototyping is possible by using Matlab, Simulink, and the Real-Time Workshop. It is shown how to automate real-time experiments and perform the interplay between on-line experiments and offline analysis. The system is able to process multiple EEG channels on-line and operates under Windows 95 in real-time on a standard PC without an additional digital signal processor (DSP) board. The BCI can be controlled over the Internet, LAN or modem. This BCI was tested on 3 subjects whose task it was to imagine either left or right hand movement. A classification accuracy between 70% and 95% could be achieved with two EEG channels after some sessions with feedback using an adaptive autoregressive (AAR) model and linear discriminant analysis (LDA).

Adolescent↗
PubMed1992

Accuracy and precision of the CellForm-Human automated sperm morphometry instrument.

OBJECTIVE: To evaluate the accuracy and precision of the CellForm-Human (CFH) automated sperm morphometry instrument (Motion Analysis Corp., Santa Rosa, CA). SETTING: Clinical and research andrology and in vitro fertilization laboratories. PATIENTS: Individuals undergoing semen evaluation and infertility work-up. RESULTS: Coefficients of variation for repeated measures of the same sperm were 1%. Coefficients of variation of normal sperm measurements were 7.4% to 12.8%, depending on the measure. Of the objects recognized as sperm by the instrument, 6.8% were debris; hence, the sperm recognition algorithms need improvement. Mean values for all CFH measures of normal sperm from specimens clinically classified as having predominantly normal, tapered, or amorphous sperm were not different; hence, the morphometry of normal sperm from normal specimens was similar to normal sperm from specimens with two different abnormalities. The instrument classified sperm as abnormal if their length or width fell outside a critical range of values recommended by the World Health Organization. Using this method, manual and CFH classification agreed unambiguously 60% of the time. When disagreement occurred, length or width marginally exceeded the range by no more than 0.1 microns. In these cases, the technician classified sperm as normal 25% of the time and classified them as abnormal 6% of the time. Because this disagreement between methods is well below the resolution of manual methods, the overall accuracy of CFH was 91% for cell type classification. CONCLUSION: At its present stage of development, the CFH instrument exceeds the accuracy and precision of most manual approaches. With improvements in sperm recognition and type classification algorithms, it could significantly improve the reliability of morphology assays in clinical and research laboratories.

Equipment Design↗
PubMed2004

A dependency model for patients with Alzheimer's disease: its validation and relationship to the costs of care--the LASER-AD Study.

BACKGROUND: Loss of independence becomes more marked as Alzheimer's disease (AD) progresses and contributes significantly to its societal and economic burden. Existing measures of functional disability focus either on basic or on instrumental activities of daily living (ADL). It would be more appropriate to combine these but, using existing assessment tools, this would involve considerable quantitative analysis. Recently, a qualitative and pragmatic system of classifying AD patients according to levels of dependency has been developed in a Belgian cohort. OBJECTIVES: To validate independently, in a UK community setting, a functional classification model of AD patients and to explore the relationship between dependency and costs of care using this model. RESEARCH DESIGN AND METHODS: Longitudinal epidemiological study of 224 AD patients. Data were collected at baseline and at 6 months on ADL, global state, cognition, behavioural dimensions, depression, quality of life and resource utilisation using validated instruments. An automatic classification algorithm was performed to allow identification of dependency clusters. The scheme was tested for validity against other simultaneously collected data including health and social care costs. The relationship between dependency and costs of care was explored using ANOVA models. RESULTS: Analysis of the ADL assessment instruments produced three ADL sub-scores by which patients could be classified into one of three disability clusters: ('non-dependent', 'non-dependent with instrumental functional disability', and 'dependent'). Good external validity of the classification scheme was demonstrated by correlation with simultaneously collected data. After a backward selection process on ANOVA model (at a 5% level), institutionalisation and the most dependent status were the most significant cost drivers. CONCLUSIONS: Qualitative classification of AD patients using dependency levels is a simple and validated approach. Applying this approach showed that institutionalisation and the most 'dependent' status were independently and significantly associated with high care cost.

Activities of Daily Living↗
PubMed2003

QSAR and classification of murine and human soluble epoxide hydrolase inhibition by urea-like compounds.

A data set of 348 urea-like compounds that inhibit the soluble epoxide hydrolase enzyme in mice and humans is examined. Compounds having IC(50) values ranging from 0.06 to >500 microM (murine) and 0.10 to >500 microM (human) are categorized as active or inactive for classification, while quantitation is performed on smaller compound subsets ranging from 0.07 to 431 microM (murine) and 0.11 to 490 microM (human). Each compound is represented by calculated structural descriptors that encode topological, geometrical, electronic, and polar surface features. Multiple linear regression (MLR) and computational neural networks (CNNs) are employed for quantitative models. Three classification algorithms, k-nearest neighbor (kNN), linear discriminant analysis (LDA), and radial basis function neural networks (RBFNN), are used to categorize compounds as active or inactive based on selected data split points. Quantitative modeling of human enzyme inhibition results in a nonlinear, five-descriptor model with root-mean-square errors (log units of IC(50) [microM]) of 0.616 (r(2) = 0.66), 0.674 (r(2) = 0.61), and 0.914 (r(2) = 0.33) for training, cross-validation, and prediction sets, respectively. The best classification results for human and murine enzyme inhibition are found using kNN. Human classification rates using a seven-descriptor model for training and prediction sets are 89.1% and 91.4%, respectively. Murine classification rates using a five-descriptor model for training and prediction sets are 91.5% and 88.6%, respectively.

Animals↗
PubMed2006

[Serum proteomic profiles of the subjects with esophageal precancerous and cancerous lesions from Linzhou, an area with high incidence of esophageal cancer in Henan Province, Northern China].

BACKGROUND & OBJECTIVE: Some molecular changes occurred in esophageal precancerous and cancerous lesions could be reflected in the serum, but the clinical application is limited because of their low sensitivity and specificity. Serum proteomic profiling is much desirable in identifying the proteins closely related to esophageal carcinogenesis. This study was to characterize the serum protein profiles of the subjects with normal esophagus, esophageal precancerous and cancerous lesions from Linzhou, the area with the highest incidence of esophageal carcinoma (EC) in Henan Province, Northern China. METHODS: Proteomic spectra were generated with surface-enhanced laser desorption/inionation-time of flight-mass spectra (SELDI-TOF-MS) and weak cation exchange (WCX2) protein chip system, and analyzed by bioinformatics like decision tree classification algorithm on a set of serum from 130 symptom-free subjects [including 63 cases with normal esophageal epithelia, 40 with basal cell hyperplasia (BCH), and 27 with dysplasia (DYS)] and 30 EC patients from Linzhou. RESULTS: One protein in BCH group with a ratio of mass to charge (M/Z) of M9 306.61 u, 1 in DYS group with a M/Z ratio of M13 765.9 u, and 2 in EC group with M/Z ratios of M2 942.15 u and M15 953.4 u were selected to build 3 decision tree classification models to identify the subjects with BCH, DYS, and EC, respectively. With these classification models, the sensitivities of identifying BCH, DYS and EC were 57.5% (23/40), 88.8% (24/27) and 96.6% (29/30), respectively, in the training sets, and 57.5% (23/40), 66.6% (18/27) and 60.0% (18/30), respectively, in the test sets; the specificities of identifying BCH, DYS and EC were 96.8% (61/63), 63.4% (40/63) and 92.0% (58/63), respectively, in the training sets, and 95.2% (60/63), 71.4% (45/63) and 84.1% (53/63), respectively, in the test sets. CONCLUSION: The protein sets with M/Z ratios of M9 306.61 u, M13 765.9 u, M2 942.15 u, and M15 953.4 u may contain promising serum biomarkers for screening the subjects with high-risk of EC.

Adult↗
PubMed2005

A novel approach for clustering proteomics data using Bayesian fast Fourier transform.

MOTIVATION: Bioinformatics clustering tools are useful at all levels of proteomic data analysis. Proteomics studies can provide a wealth of information and rapidly generate large quantities of data from the analysis of biological specimens. The high dimensionality of data generated from these studies requires the development of improved bioinformatics tools for efficient and accurate data analyses. For proteome profiling of a particular system or organism, a number of specialized software tools are needed. Indeed, significant advances in the informatics and software tools necessary to support the analysis and management of these massive amounts of data are needed. Clustering algorithms based on probabilistic and Bayesian models provide an alternative to heuristic algorithms. The number of clusters (diseased and non-diseased groups) is reduced to the choice of the number of components of a mixture of underlying probability. The Bayesian approach is a tool for including information from the data to the analysis. It offers an estimation of the uncertainties of the data and the parameters involved. RESULTS: We present novel algorithms that can organize, cluster and derive meaningful patterns of expression from large-scaled proteomics experiments. We processed raw data using a graphical-based algorithm by transforming it from a real space data-expression to a complex space data-expression using discrete Fourier transformation; then we used a thresholding approach to denoise and reduce the length of each spectrum. Bayesian clustering was applied to the reconstructed data. In comparison with several other algorithms used in this study including K-means, (Kohonen self-organizing map (SOM), and linear discriminant analysis, the Bayesian-Fourier model-based approach displayed superior performances consistently, in selecting the correct model and the number of clusters, thus providing a novel approach for accurate diagnosis of the disease. Using this approach, we were able to successfully denoise proteomic spectra and reach up to a 99% total reduction of the number of peaks compared to the original data. In addition, the Bayesian-based approach generated a better classification rate in comparison with other classification algorithms. This new finding will allow us to apply the Fourier transformation for the selection of the protein profile for each sample, and to develop a novel bioinformatic strategy based on Bayesian clustering for biomarker discovery and optimal diagnosis.

Algorithms↗
PubMed1996

An on-line automated sleep-wake classification system for laboratory animals.

A computerized sleep-wake classification program is presented that is capable of classifying sleep-wake states on-line in four animals simultaneously. Every 10 s the classification algorithm assigns sleep-wake states on the basis of the power spectrum of an EEG signal and the standard deviation of an EMG signal. The system was developed specifically for long-term circadian rhythm experiments and can run uninterrupted for an unlimited period of time. To validate the system a total of 5760 sleep-wake epochs from 4 rats was classified both visually and by computer. The results show a good agreement between visual and computer automated (kappa = 0.770) sleep-wake scoring. Therefore, the presented sleep-wake classification system can be regarded as a reliable time saving alternative for visual scoring in sleep research.

Animals↗
PubMed1997

Sudden unexplained death in epilepsy: observations from a large clinical development program.

PURPOSE: The present study was conducted to determine the rate of sudden unexplained death in epilepsy (SUDEP) in a well-defined cohort of patients included in the lamotrigine (LTG) clinical development database. METHODS: A panel of scientists experienced in the area of SUDEP was assembled and provided with case summaries on all deaths (n = 45) reported during the initial clinical development of LTG. The panel developed a set of criteria for classifying cases as SUDEP (definite or highly probable), possible SUDEP, or non-SUDEP. This classification algorithm was then applied to the LTG cases, and SUDEP rates were calculated using patient-years of exposure as the denominator. RESULTS: At the time of the study, 4,700 patients (5,747 patient-years of exposure) were included in the worldwide LTG clinical trials database. In this cohort, 45 deaths were reported. Eighteen were judged by the panel to be SUDEP, 6 were defined as possible SUDEP, 20 were judged to be due to other causes (non-SUDEP), and 1 lacked sufficient data from which to make a classification. The overall SUDEP rate (definite/ highly probable SUDEP and possible SUDEP combined) was calculated to be 3.5 in 1,000 patient-years of exposure to LTG. CONCLUSIONS: The rate of SUDEP in this cohort of patients was comparable to the rate that would be expected in young adults with severe epilepsy (the subgroup of patients believed to be at highest risk of SUDEP). The data suggest that the rate of SUDEP in the LTG clinical development program is a function of the clinical trial population and is unrelated to drug treatment.

Adolescent↗
PubMed2003

Use of Fourier-transform infrared spectroscopy for typing of Candida albicans strains isolated in intensive care units.

Comparative studies of Candida albicans strains are essential for proving cross-infections in epidemiological investigations. Typing of C. albicans strains is mainly based on genotypic methods. Fourier-transform infrared (FTIR) spectroscopy is described in this study as a novel phenotypic approach to the typing of C. albicans. The first step in the approach was the standardization of sample preparation (culture conditions and sampling parameters) and acquisition and classification parameters (spectral acquisition, spectral window selection, classification algorithm, and heterogeneity threshold). The second step consisted of validating the established parameters with a set of 79 strains of C. albicans isolated over 4 months from nine patients hospitalized in two intensive care units. Strains were isolated from multiple anatomical sites with repeated sampling. FTIR spectroscopy results were compared to randomly amplified polymorphic DNA (RAPD) results; this analysis showed that the amplification patterns of strains isolated from a given patient were identical and that different patients had different profiles. FTIR spectroscopy data were analyzed by hierarchical clustering performed with the second-derivative spectra. This classification revealed nine groups, one per patient. Only one spectrum out of 79 was misclassified by the FTIR spectroscopy method. RAPD and FTIR spectroscopy results were in good agreement, showing that, when nosocomial candidiasis transmission is suspected and urgent information is needed, this technique may be useful as a quick identification tool to give solid clues before confirmation by a genotypic method.

Adult↗
PubMedDate not supplied

Evaluation of a "lexically assign, logically refine" strategy for semi-automated integration of overlapping terminologies.

OBJECTIVE: To evaluate a "lexically assign, logically refine" (LALR) strategy for merging overlapping healthcare terminologies. This strategy combines description logic classification with lexical techniques that propose initial term definitions. The lexically suggested initial definitions are manually refined by domain experts to yield description logic definitions for each term in the overlapping terminologies of interest. Logic-based techniques are then used to merge defined terms. METHODS: A LALR strategy was applied to 7,763 LOINC and 2,050 SNOMED procedure terms using a common set of defining relationships taken from the LOINC data model. Candidate value restrictions were derived by lexically comparing the procedure's name with other terms contained in the reference SNOMED topography, living organism, function, and chemical axes. These candidate restrictions were reviewed by a domain expert, transformed into terminologic definitions for each of the terms, and then algorithmically classified. RESULTS: The authors successfully defined 5,724 (73%) LOINC and 1,151 (56%) SNOMED procedure terms using a LALR strategy. Algorithmic classification of the defined concepts resulted in an organization mirroring that of the reference hierarchies. The classification techniques appropriately placed more detailed LOINC terms underneath the corresponding SNOMED terms, thus forming a complementary relationship between the LOINC and SNOMED terms. DISCUSSION: LALR is a successful strategy for merging overlapping terminologies in a test case where both terminologies can be defined using the same defining relationships, and where value restrictions can be drawn from a single reference hierarchy. Those concepts not having lexically suggested value restrictions frequently indicate gaps in the reference hierarchy.

Algorithms↗
PubMed2003

Remote detection of heated ethanol plumes by airborne passive Fourier transform infrared spectrometry.

Methodology is developed for the automated detection of heated plumes of ethanol vapor with airborne passive Fourier transform infrared spectrometry. Positioned in a fixed-wing aircraft in a downward-looking mode, the spectrometer is used to detect ground sources of ethanol vapor from an altitude of 2000-3000 ft. Challenges to the use of this approach for the routine detection of chemical plumes include (1) the presence of a constantly changing background radiance as the aircraft flies, (2) the cost and complexity of collecting the data needed to train the classification algorithms used in implementing the plume detection, and (3) the need for rapid interferogram scans to minimize the ground area viewed per scan. To address these challenges, this work couples a novel ground-based data collection and training protocol with the use of signal processing and pattern recognition methods based on short sections of the interferogram data collected by the spectrometer. In the data collection, heated plumes of ethanol vapor are released from a portable emission stack and viewed by the spectrometer from ground level against a synthetic background designed to simulate a terrestrial radiance source. Classifiers trained with these data are subsequently tested with airborne data collected over a period of 2.5 years. Two classifier architectures are compared in this work: support vector machines (SVM) and piecewise linear discriminant analysis (PLDA). When applied to the airborne test data, the SVM classifiers perform best, failing to detect ethanol in only 8% of the cases in which it is present. False detections occur at a rate of less than 0.5%. The classifier performs well in spite of differences between the backgrounds associated with the ground-based and airborne data collections and the instrumental drift arising from the long time span of the data collection. Further improvements in classification performance are judged to require increased sophistication in the ground-based data collection in order to provide a better match to the infrared backgrounds observed from the air.

Air Pollutants↗
PubMed1998

Computer-aided diagnoses: automatic detection of lung nodules.

This work describes a computational scheme for automatic detection of suspected lung nodules in a chest radiograph. A knowledge-based system extracts the lung masks over which we will apply the nodule detection process. First we obtain the normalized cross-correlation image. Next we detect suspicious regions by assuming a threshold. We examine the suspicious regions using a variable threshold which results in the growth of the suspicious areas and an increase in false positives. We reduce the large number of false positives by applying the facet model to the suspicious regions of the image. An algorithmic classification process gives a confidence factor that a suspicious region is a nodule. Five chest images containing 30 known nodules were used as a training set. We evaluated the system by analyzing 30 chest images with 40 confirmed nodules of varying contrast and size located in various parts of the lungs. The system detected 100% of the nodules with a mean of six false positives per image. The accuracy and specificity were 96%.

Algorithms↗