Application and comparison of classification algorithms for recognition of Alzheimer's disease in electrical brain activity (EEG)

The early detection of subjects with probable Alzheimer's disease (AD) is crucial for effective appliance of treatment strategies. Here we explored the ability of a multitude of linear and non-linear classification algorithms to discriminate between the electroencephalograms (EEGs) of patients with varying degree of AD and their age-matched control subjects. Absolute and relative spectral power, distribution of spectral power, and measures of spatial synchronization were calculated from recordings of resting eyes-closed continuous EEGs of 45 healthy controls, 116 patients with mild AD and 81 patients with moderate AD, recruited in two different centers (Stockholm, New York). The applied classification algorithms were: principal component linear discriminant analysis (PC LDA), partial least squares LDA (PLS LDA), principal component logistic regression (PC LR), partial least squares logistic regression (PLS LR), bagging, random forest, support vector machines (SVM) and feed-forward neural network. Based on 10-fold cross-validation runs it could be demonstrated that even tough modern computer-intensive classification algorithms such as random forests, SVM and neural networks show a slight superiority, more classical classification algorithms performed nearly equally well. Using random forests classification a considerable sensitivity of up to 85% and a specificity of 78%, respectively for the test of even only mild AD patients has been reached, whereas for the comparison of moderate AD vs. controls, using SVM and neural networks, values of 89% and 88% for sensitivity and specificity were achieved. Such a remarkable performance proves the value of these classification algorithms for clinical diagnostics.

Classification Using Generalized Partial Least Squares

The advances in computational biology have made simultaneous monitoring of thousands of features possible. The high throughput technologies not only bring about a much richer information context in which to study various aspects of gene functions but they also present challenge of analyzing data with large number of covariates and few samples. As an integral part of machine learning, classification of samples into two or more categories is almost always of interest to scientists. In this paper, we address the question of classification in this setting by extending partial least squares (PLS), a popular dimension reduction tool in chemometrics, in the context of generalized linear regression based on a previous approach, Iteratively ReWeighted Partial Least Squares, i.e. IRWPLS (Marx, 1996). We compare our results with two-stage PLS (Nguyen and Rocke, 2002A; Nguyen and Rocke, 2002B) and other classifiers. We show that by phrasing the problem in a generalized linear model setting and by applying bias correction to the likelihood to avoid (quasi)separation, we often get lower classification error rates.

Model Evaluation Based on the Distribution of Estimated Absolute Prediction Error

The construction of a reliable, practically useful prediction rule for future response is heavily dependent on the "adequacy" of the fitted regression model. In this article, we consider the absolute prediction error, the expected value of the absolute difference between the future and predicted responses, as the model evaluation criterion. This prediction error is easier to interpret than the average squared error and is equivalent to the mis-classification error for the binary outcome. We show that the distributions of the apparent error and its cross-validation counterparts are approximately normal even under a misspecified fitted model. When the prediction rule is "unsmooth", the variance of the above normal distribution can be estimated well via a perturbation-resampling method. We also show how to approximate the distribution of the difference of the estimated prediction errors from two competing models. With two real examples, we demonstrate that the resulting interval estimates for prediction errors provide much more information about model adequacy than the point estimates alone.

Clinical forensic radiology in strangulation victims: forensic expertise based on magnetic resonance imaging (MRI) findings

Based on only one objective and several subjective signs, the forensic classification of strangulation incidents concerning their life-threatening quality can be problematic. Reflecting that it is almost impossible to detect internal injuries of the neck with the standard forensic external examination, we examined 14 persons who have survived manual and ligature strangulation or forearm choke holds using MRI technique (1.5-T scanner). Two clinical radiologists evaluated the neck findings independently. The danger to life was evaluated based on the "classical" external findings alone and in addition to the radiological data. We observed hemorrhaging in the subcutaneous fatty tissue of the neck in ten cases. Other frequent findings were hemorrhages of the neck and larynx muscles, the lymph nodes, the pharynx, and larynx soft tissues. Based on the classical forensic strangulation findings with MRI, eight of the cases were declared as life-endangering incidents, four of them without the presence of petechial hemorrhage but with further signs of impaired brain function due to hypoxia. The accuracy of future forensic classification of the danger to life will probably be increased when it is based not only on one objective and several subjective signs but also on the evidence of inner neck injuries. However, further prospective studies including larger cohorts are necessary to clarify the value of the inner neck injuries in the forensic classification of surviving strangulation victims.

A novel 2D-based approach to the discovery of candidate substrates for the metalloendopeptidase meprin

In the past, protease-substrate finding proved to be rather haphazard and was executed by in vitro cleavage assays using singly selected targets. In the present study, we report the first protease proteomic approach applied to meprin, an astacin-like metalloendopeptidase, to determine physiological substrates in a cell-based system of Madin-Darby canine kidney epithelial cells. A simple 2D IEF/SDS/PAGE-based image analysis procedure was designed to find candidate substrates in conditioned media of Madin-Darby canine kidney cells expressing meprin in zymogen or in active form. The method enabled the discovery of hitherto unknown meprin substrates with shortened (non-trypsin-generated) N- and C-terminally truncated cleavage products in peptide fragments upon LC-MS/MS analysis. Of 22 (17 nonredundant) candidate substrates identified, the proteolytic processing of vinculin, lysyl oxidase, collagen type V and annexin A1 was analysed by means of immunoblotting validation experiments. The classification of substrates into functional groups may propose new functions for meprins in the regulation of cell homeostasis and the extracellular environment, and in innate immunity, respectively.

Development of a physically-based method for delineation of hydrologically homogeneous regions and flood quantile estimation in ungauged basins via the index flood method

Regional flood frequency techniques are commonly used to estimate flood quantiles when flood data is unavailable or the record length at an individual gauging station is insufficient for reliable analyses. These methods compensate for limited or unavailable data by pooling data from nearby gauged sites. This requires the delineation of hydrologically homogeneous regions in which the flood regime is sufficiently similar to allow the spatial transfer of information. It is generally accepted that hydrologic similarity results from similar physiographic characteristics, and thus these characteristics can be used to delineate regions and classify ungauged sites. However, as currently practiced, the delineation is highly subjective and dependent on the similarity measures and classification techniques employed. A standardized procedure for delineation of hydrologically homogeneous regions is presented herein. Key aspects are a new statistical metric to identify physically discordant sites, and the identification of an appropriate set of physically based measures of extreme hydrological similarity. A combination of multivariate statistical techniques applied to multiple flood statistics and basin characteristics for gauging stations in the Southeastern U.S. revealed that basin slope, elevation, and soil drainage largely determine the extreme hydrological behavior of a watershed. Use of these characteristics as similarity measures in the standardized approach for region delineation yields regions which are more homogeneous and more efficient for quantile estimation at ungauged sites than those delineated using alternative physically-based procedures typically employed in practice. The proposed methods and key physical characteristics are also shown to be efficient for region delineation and quantile development in alternative areas composed of watersheds with statistically different physical composition. In addition, the use of aggregated values of key watershed characteristics was found to be sufficient for the regionalization of flood data; the added time and computational effort required to derive spatially distributed watershed variables does not increase the accuracy of quantile estimators for ungauged sites. This dissertation also presents a methodology by which flood quantile estimates in Haiti can be derived using relationships developed for data rich regions of the U.S. As currently practiced, regional flood frequency techniques can only be applied within the predefined area used for model development. However, results presented herein demonstrate that the regional flood distribution can successfully be extrapolated to areas of similar physical composition located beyond the extent of that used for model development provided differences in precipitation are accounted for and the site in question can be appropriately classified within a delineated region.

Comparing remote sensing and ground-based methods of quantifying cover type and carbon storage in an urban forest

Understanding the canopy cover of an urban environment leads to better estimates of carbon storage and more informed management decisions by urban foresters. The most commonly used method for assessing urban forest cover type extent is ground surveys, which can be both timeconsuming and expensive. The analysis of aerial photos is an alternative method that is faster, cheaper, and can cover a larger number of sites, but may be less accurate. The objectives of this paper were (1) to compare three methods of cover type assessment for Los Angeles, CA: handdelineation of aerial photos in ArcMap, supervised classification of aerial photos in ERDAS Imagine, and ground-collected data using the Urban Forest Effects (UFORE) model protocol; (2) to determine how well remote sensing methods estimate carbon storage as predicted by the UFORE model; and (3) to explore the influence of tree diameter and tree density on carbon storage estimates. Four major cover types (bare ground, fine vegetation, coarse vegetation, and impervious surfaces) were determined from 348 plots (0.039 ha each) randomly stratified according to land-use. Hand-delineation was better than supervised classification at predicting ground-based measurements of cover type and UFORE model-predicted carbon storage. Most error in supervised classification resulted from shadow, which was interpreted as unknown cover type. Neither tree diameter or tree density per plot significantly affected the relationship between carbon storage and canopy cover. The efficiency of remote sensing rather than in situ data collection allows urban forest managers the ability to quickly assess a city and plan accordingly while also preserving their often-limited budget.

Salivary stones and stenosis. A comprehensive classification

INTRODUCTION: Sialendoscopy and sialoMRI enables diagnosis of salivary gland obstructive pathologies, such as lithiasis, stenosis, and dilatations. Therefore, a classification of these pathologies is needed, allowing large series comparisons, for better diagnosis and treatment of salivary pathologies. MATERIAL AND METHODS: With help from people from the European Sialendoscopy Training Center (ESTC), the results of sialographies, sialoMRI and sialendoscopies, a comprehensive classification of obstructive salivary pathologies is described, based on the absence or presence of lithiasis (L), stenosis (S), and dilatation (D) ("LSD" classification). DISCUSSION: It appears that a classification of salivary gland obstructive pathologies should be described. We hope it will be widely used and of course criticized to be improved and to compare the results of salivary gland diagnostic methods, such as sialography and sialendoscopy, and also the results and indications for salivary gland therapeutic methods, such as lithotripsy, sialendoscopy, and/or open surgery.

A MEMS Light Modulator Based on Diffractive Nanohole Gratings

We present the design, fabrication, and testing of a microelectromechanical systems (MEMS) light modulator based on pixels patterned with periodic nanohole arrays. Flexure-suspended silicon pixels are patterned with a two dimensional array of 150 nm diameter nanoholes using nanoimprint lithography. A top glass plate assembled above the pixel array is used to provide a counter electrode for electrostatic actuation. The nanohole pattern is designed so that normally-incident light is coupled into an in-plane grating resonance, resulting in an optical stop-band at a desired wavelength. When the pixel is switched into contact with the top plate, the pixel becomes highly reflective. A 3:1 contrast ratio at the resonant wavelength is demonstrated for gratings patterned on bulk Si substrates. The switching time is 0.08 ms and the switching voltage is less than 15V.

A disease model for wheezing disorders in preschool children based on clinicians' perceptions

BACKGROUND: Wheezing disorders in childhood vary widely in clinical presentation and disease course. During the last years, several ways to classify wheezing children into different disease phenotypes have been proposed and are increasingly used for clinical guidance, but validation of these hypothetical entities is difficult. METHODOLOGY/PRINCIPAL FINDINGS: The aim of this study was to develop a testable disease model which reflects the full spectrum of wheezing illness in preschool children. We performed a qualitative study among a panel of 7 experienced clinicians from 4 European countries working in primary, secondary and tertiary paediatric care. In a series of questionnaire surveys and structured discussions, we found a general consensus that preschool wheezing disorders consist of several phenotypes, with a great heterogeneity of specific disease concepts between clinicians. Initially, 24 disease entities were described among the 7 physicians. In structured discussions, these could be narrowed down to three entities which were linked to proposed mechanisms: a) allergic wheeze, b) non-allergic wheeze due to structural airway narrowing and c) non-allergic wheeze due to increased immune response to viral infections. This disease model will serve to create an artificial dataset that allows the validation of data-driven multidimensional methods, such as cluster analysis, which have been proposed for identification of wheezing phenotypes in children. CONCLUSIONS/SIGNIFICANCE: While there appears to be wide agreement among clinicians that wheezing disorders consist of several diseases, there is less agreement regarding their number and nature. A great diversity of disease concepts exist but a unified phenotype classification reflecting underlying disease mechanisms is lacking. We propose a disease model which may help guide future research so that proposed mechanisms are measured at the right time and their role in disease heterogeneity can be studied.

Computer-aided diagnosis of carotid atherosclerosis based on ultrasound image statistics, laws' texture and neural networks

Quantitative characterisation of carotid atherosclerosis and classification into symptomatic or asymptomatic is crucial in planning optimal treatment of atheromatous plaque. The computer-aided diagnosis (CAD) system described in this paper can analyse ultrasound (US) images of carotid artery and classify them into symptomatic or asymptomatic based on their echogenicity characteristics. The CAD system consists of three modules: a) the feature extraction module, where first-order statistical (FOS) features and Laws' texture energy can be estimated, b) the dimensionality reduction module, where the number of features can be reduced using analysis of variance (ANOVA), and c) the classifier module consisting of a neural network (NN) trained by a novel hybrid method based on genetic algorithms (GAs) along with the back propagation algorithm. The hybrid method is able to select the most robust features, to adjust automatically the NN architecture and to optimise the classification performance. The performance is measured by the accuracy, sensitivity, specificity and the area under the receiver-operating characteristic (ROC) curve. The CAD design and development is based on images from 54 symptomatic and 54 asymptomatic plaques. This study demonstrates the ability of a CAD system based on US image analysis and a hybrid trained NN to identify atheromatous plaques at high risk of stroke.

Content Classification of Multimedia Documents using Partitions of Low-Level Features

Audio-visual documents obtained from German TV news are classified according to the IPTC topic categorization scheme. To this end usual text classification techniques are adapted to speech, video, and non-speech audio. For each of the three modalities word analogues are generated: sequences of syllables for speech, “video words” based on low level color features (color moments, color correlogram and color wavelet), and “audio words” based on low-level spectral features (spectral envelope and spectral flatness) for non-speech audio. Such audio and video words provide a means to represent the different modalities in a uniform way. The frequencies of the word analogues represent audio-visual documents: the standard bag-of-words approach. Support vector machines are used for supervised classification in a 1 vs. n setting. Classification based on speech outperforms all other single modalities. Combining speech with non-speech audio improves classification. Classification is further improved by supplementing speech and non-speech audio with video words. Optimal F-scores range between 62% and 94% corresponding to 50% - 84% above chance. The optimal combination of modalities depends on the category to be recognized. The construction of audio and video words from low-level features provide a good basis for the integration of speech, non-speech audio and video.

Classification of the taxon 2 and taxon 3 complex of Bisgaard within Gallibacterium and description of Gallibacterium melopsittaci sp. nov., Gallibacterium trehalosifermentans sp. nov. and Gallibacterium salpingitidis sp. nov

This investigation was based on 23 isolates from several European countries collected over the past 30 years, and included characterization of all isolates. Published data on amplified fragment length polymorphism typing of isolates representing all biovars as well as protein profiles were used to select strains that were then further characterized by polyamine profiling and sequencing of 16S rRNA, infB, rpoB and recN genes. Comparison of 16S rRNA gene sequences revealed a monophyletic group within the avian 16S rRNA group of the Pasteurellaceae, which currently includes the genera Avibacterium, Gallibacterium and Volucribacter. Five monophyletic subgroups related to Gallibacterium anatis were recognized by 16S rRNA, rpoB, infB and recN gene sequence comparisons. Whole-genome similarity between strains of the five subgroups and the type strain of G. anatis calculated from recN sequences allowed us to classify them within the genus Gallibacterium. In addition, phenotypic data including biochemical traits, protein profiling and polyamine patterns clearly indicated that these taxa are related. Major phenotypic diversity was observed for 16S rRNA gene sequence groups. Furthermore, comparison of whole-genome similarities, phenotypic data and published data on amplified fragment length polymorphism and protein profiling revealed that each of the five groups present unique properties that allow the proposal of three novel species of Gallibacterium, for which we propose the names Gallibacterium melopsittaci sp. nov. (type strain F450(T) =CCUG 36331(T) =CCM 7538(T)), Gallibacterium trehalosifermentans sp. nov. (type strain 52/S3/90(T) =CCUG 55631(T) =CCM 7539(T)) and Gallibacterium salpingitidis sp. nov. (type strain F150(T) =CCUG 15564(T) =CCUG 36325(T) =NCTC 11414(T)), a novel genomospecies 3 of Gallibacterium and an unnamed taxon (group V). An emended description of the genus Gallibacterium is also presented.

IDH/MGMT-driven molecular classification of low-grade glioma is a strong predictor for long-term survival

BACKGROUND Low-grade gliomas (LGGs) are rare brain neoplasms, with survival spanning up to a few decades. Thus, accurate evaluations on how biomarkers impact survival among patients with LGG require long-term studies on samples prospectively collected over a long period. METHODS The 210 adult LGGs collected in our databank were screened for IDH1 and IDH2 mutations (IDHmut), MGMT gene promoter methylation (MGMTmet), 1p/19q loss of heterozygosity (1p19qloh), and nuclear TP53 immunopositivity (TP53pos). Multivariate survival analyses with multiple imputation of missing data were performed using either histopathology or molecular markers. Both models were compared using Akaike's information criterion (AIC). The molecular model was reduced by stepwise model selection to filter out the most critical predictors. A third model was generated to assess for various marker combinations. RESULTS Molecular parameters were better survival predictors than histology (ΔAIC = 12.5, P< .001). Forty-five percent of studied patients died. MGMTmet was positively associated with IDHmut (P< .001). In the molecular model with marker combinations, IDHmut/MGMTmet combined status had a favorable impact on overall survival, compared with IDHwt (hazard ratio [HR] = 0.33, P< .01), and even more so the triple combination, IDHmut/MGMTmet/1p19qloh (HR = 0.18, P< .001). Furthermore, IDHmut/MGMTmet/TP53pos triple combination was a significant risk factor for malignant transformation (HR = 2.75, P< .05). CONCLUSION By integrating networks of activated molecular glioma pathways, the model based on genotype better predicts prognosis than histology and, therefore, provides a more reliable tool for standardizing future treatment strategies.