Lars Kai Hansen

Neural network ensembles

Several means for improving the performance and training of neural networks for classification are proposed. Crossvalidation is used as a tool for optimizing network parameters and architecture. It is shown that the remaining residual generalization error can be reduced by invoking ensembles of similar networks. >

On Clustering fMRI Time Series

Analysis of fMRI time series is often performed by extracting one or more parameters for the individual voxels. Methods based, e.g., on various statistical tests are then used to yield parameters corresponding to probability of activation or activation strength. However, these methods do not indicate whether sets of voxels are activated in a similar way or in different ways. Typically, delays between two activated signals are not identified. In this article, we use clustering methods to detect similarities in activation between voxels. We employ a novel metric that measures the similarity between the activation stimulus and the fMRI signal. We present two different clustering algorithms and use them to identify regions of similar activations in an fMRI experiment involving a visual stimulus.

Independent component analysis of functional MRI: what is signal and what is noise?

Many sources of fluctuation contribute to the functional magnetic resonance imaging (fMRI) signal, complicating attempts to infer those changes that are truly related to brain activation. Unlike methods of analysis of fMRI data that test the time course of each voxel against a hypothesized waveform, data-driven methods, such as independent component analysis and clustering, attempt to find common features within the data. This exploratory approach can be revealing when the brain activation is difficult to predict beforehand, such as with complex stimuli and internal shifts of activation that are not time-locked to an easily specified sensory or motor event. These methods can be further improved by incorporating prior knowledge regarding the temporal and spatial extent of brain activation.

The Quantitative Evaluation of Functional Neuroimaging Experiments: The NPAIRS Data Analysis Framework

We introduce a data-analysis framework and performance metrics for evaluating and optimizing the interaction between activation tasks, experimental designs, and the methodological choices and tools for data acquisition, preprocessing, data analysis, and extraction of statistical parametric maps (SPMs). Our NPAIRS (nonparametric prediction, activation, influence, and reproducibility resampling) framework provides an alternative to simulations and ROC curves by using real PET and fMRI data sets to examine the relationship between prediction accuracy and the signal-to-noise ratios (SNRs) associated with reproducible SPMs. Using cross-validation resampling we plot training-test set predictions of the experimental design variables (e.g., brain-state labels) versus reproducibility SNR metrics for the associated SPMs. We demonstrate the utility of this framework across the wide range of performance metrics obtained from [(15)O]water PET studies of 12 age- and sex-matched data sets performing different motor tasks (8 subjects/set). For the 12 data sets we apply NPAIRS with both univariate and multivariate data-analysis approaches to: (1) demonstrate that this framework may be used to obtain reproducible SPMs from any data-analysis approach on a common Z-score scale (rSPM[Z]); (2) demonstrate that the histogram of a rSPM[Z] image may be modeled as the sum of a data-analysis-dependent noise distribution and a task-dependent, Gaussian signal distribution that scales monotonically with our reproducibility performance metric; (3) explore the relation between prediction and reproducibility performance metrics with an emphasis on bias-variance tradeoffs for flexible, multivariate models; and (4) measure the broad range of reproducibility SNRs and the significant influence of individual subjects. A companion paper describes learning curves for four of these 12 data sets, which describe an alternative mutual-information prediction metric and NPAIRS reproducibility as a function of training-set sizes from 2 to 18 subjects. We propose the NPAIRS framework as a validation tool for testing and optimizing methodological choices and tools in functional neuroimaging.

Parallel Factor Analysis as an exploratory tool for wavelet transformed event-related EEG

In the decomposition of multi-channel EEG signals, principal component analysis (PCA) and independent component analysis (ICA) have widely been used. However, as both methods are based on handling two-way data, i.e. two-dimensional matrices, multi-way methods might improve the interpretation of frequency transformed multi-channel EEG of channel x frequency x time data. The multi-way decomposition method Parallel Factor (PARAFAC), also named Canonical Decomposition (CANDECOMP), was recently used to decompose the wavelet transformed ongoing EEG of channel x frequency x time (Miwakeichi, F., Martinez-Montes, E., Valdes-Sosa, P.A., Nishiyama, N., Mizuhara, H., Yamaguchi, Y., 2004. Decomposing EEG data into space-time-frequency components using parallel factor analysis. Neuroimage 22, 1035-1045). In this article, PARAFAC is used for the first time to decompose wavelet transformed event-related EEG given by the inter-trial phase coherence (ITPC) encompassing ANOVA analysis of differences between conditions and 5-way analysis of channel x frequency x time x subject x condition. A flow chart is presented on how to perform data exploration using the PARAFAC decomposition on multi-way arrays. This includes (A) channel x frequency x time 3-way arrays of F test values from a repeated measures analysis of variance (ANOVA) between two stimulus conditions; (B) subject-specific 3-way analyses; and (C) an overall 5-way analysis of channel x frequency x time x subject x condition. The PARAFAC decompositions were able to extract the expected features of a previously reported ERP paradigm: namely, a quantitative difference of coherent occipital gamma activity between conditions of a visual paradigm. Furthermore, the method revealed a qualitative difference which has not previously been reported. The PARAFAC decomposition of the 3-way array of ANOVA F test values clearly showed the difference of regions of interest across modalities, while the 5-way analysis enabled visualization of both quantitative and qualitative differences. Consequently, PARAFAC is a promising data exploratory tool in the analysis of the wavelets transformed event-related EEG.

Growth-rate regulated genes have profound impact on interpretation of transcriptome profiling in Saccharomyces cerevisiae

BackgroundGrowth rate is central to the development of cells in all organisms. However, little is known about the impact of changing growth rates. We used continuous cultures to control growth rate and studied the transcriptional program of the model eukaryote Saccharomyces cerevisiae, with generation times varying between 2 and 35 hours.ResultsA total of 5930 transcripts were identified at the different growth rates studied. Consensus clustering of these revealed that half of all yeast genes are affected by the specific growth rate, and that the changes are similar to those found when cells are exposed to different types of stress (>80% overlap). Genes with decreased transcript levels in response to faster growth are largely of unknown function (>50%) whereas genes with increased transcript levels are involved in macromolecular biosynthesis such as those that encode ribosomal proteins. This group also covers most targets of the transcriptional activator RAP1, which is also known to be involved in replication. A positive correlation between the location of replication origins and the location of growth-regulated genes suggests a role for replication in growth rate regulation.ConclusionOur data show that the cellular growth rate has great influence on transcriptional regulation. This, in turn, implies that one should be cautious when comparing mutants with different growth rates. Our findings also indicate that much of the regulation is coordinated via the chromosomal location of the affected genes, which may be valuable information for the control of heterologous gene expression in metabolic engineering.

Detection of skin cancer by classification of Raman spectra

Skin lesion classification based on in vitro Raman spectroscopy is approached using a nonlinear neural network classifier. The classification framework is probabilistic and highly automated. The framework includes a feature extraction for Raman spectra and a fully adaptive and robust feedforward neural network classifier. Moreover, classification rules learned by the neural network may be extracted and evaluated for reproducibility, making it possible to explain the class assignment. The classification performance for the present data set, involving 222 cases and five lesion types, was 80.5%/spl plusmn/5.3% correct classification of malignant melanoma, which is similar to that of trained dermatologists based on visual inspection. The skin cancer basal cell carcinoma has a classification rate of 95.8%/spl plusmn/2.7%, which is excellent. The overall classification rate of skin lesions is 94.8%/spl plusmn/3.0%. Spectral regions, which are important for network classification, are demonstrated to reproduce. Small distinctive bands in the spectrum, corresponding to specific lipids and proteins, are shown to hold the discriminating information which the network uses to diagnose skin lesions.

Generalizable patterns in neuroimaging: how many principal components?

Generalization can be defined quantitatively and can be used to assess the performance of principal component analysis (PCA). The generalizability of PCA depends on the number of principal components retained in the analysis. We provide analytic and test set estimates of generalization. We show how the generalization error can be used to select the number of principal components in two analyses of functional magnetic resonance imaging activation sets.

Defining a Local Arterial Input Function for Perfusion MRI Using Independent Component Analysis

Quantification of cerebral blood flow (CBF) using dynamic‐susceptibility contrast MRI relies on the deconvolution of the arterial input function (AIF), which is commonly estimated from the signal changes in a major artery. However, it has been shown that the presence of bolus delay/dispersion between the artery and the tissue of interest can be a significant source of error. These effects could be minimized if a local AIF were used, although the measurement of a local AIF can be problematic. This work describes a new methodology to define a local AIF using independent component analysis (ICA). The methodology was tested on data from patients with various cerebrovascular abnormalities and compared to the conventional approach of using a global AIF. The new methodology produced higher CBF and shorter mean transit time values (compared to the global AIF case) in areas with distorted AIFs, suggesting that the effects of delay/dispersion are minimized. The minimization of these effects using the calculated local AIF should lead to a more accurate quantification of CBF, which can have important implications for diagnosis and management of patients with cerebral ischemia. Magn Reson Med 52:789–797, 2004.

Mean-field approaches to independent component analysis

We develop mean-field approaches for probabilistic independent component analysis (ICA). The sources are estimated from the mean of their posterior distribution and the mixing matrix (and noise level) is estimated by maximum a posteriori (MAP). The latter requires the computation of (a good approximation to) the correlations between sources. For this purpose, we investigate three increasingly advanced mean-field methods: the variational (also known as naive mean field) approach, linear response corrections, and an adaptive version of the Thouless, Anderson and Palmer (1977) (TAP) mean-field approach, which is due to Opper and Winther (2001). The resulting algorithms are tested on a number of problems. On synthetic data, the advanced mean-field approaches are able to recover the correct mixing matrix in cases where the variational mean-field theory fails. For handwritten digits, sparse encoding is achieved using nonnegative source and mixing priors. For speech, the mean-field method is able to separate in the underdetermined (overcomplete) case of two sensors and three sources. One major advantage of the proposed method is its generality and algorithmic simplicity. Finally, we point out several possible extensions of the approaches developed here.