Improvements on ICA mixture models for image pre-processing and segmentation

Today several different unsupervised classification algorithms are commonly used to cluster similar patterns in a data set based only on its statistical properties. Specially in image data applications, self-organizing methods for unsupervised classification have been successfully applied for clustering pixels or group of pixels in order to perform segmentation tasks. The first important contribution of this paper refers to the development of a self-organizing method for data classification, named Enhanced Independent Component Analysis Mixture Model (EICAMM), which was built by proposing some modifications in the Independent Component Analysis Mixture Model (ICAMM). Such improvements were proposed by considering some of the model limitations as well as by analyzing how it should be improved in order to become more efficient. Moreover, a pre-processing methodology was also proposed, which is based on combining the Sparse Code Shrinkage (SCS) for image denoising and the Sobel edge detector. In the experiments of this work, the EICAMM and other self-organizing models were applied for segmenting images in their original and pre-processed versions. A comparative analysis showed satisfactory and competitive image segmentation results obtained by the proposals presented herein. (C) 2008 Published by Elsevier B.V.

White matter microstructure underlying default mode network connectivity in the human brain

Resting state functional magnetic resonance imaging (fMRI) reveals a distinct network of correlated brain function representing a default mode state of the human brain The underlying structural basis of this functional connectivity pattern is still widely unexplored We combined fractional anisotropy measures of fiber tract integrity derived from diffusion tensor imaging (DTI) and resting state fMRI data obtained at 3 Tesla from 20 healthy elderly subjects (56 to 83 years of age) to determine white matter microstructure e 7 underlying default mode connectivity We hypothesized that the functional connectivity between the posterior cingulate and hippocampus from resting state fMRI data Would be associated with the white matter microstructure in the cingulate bundle and fiber tracts connecting posterior cingulate gyrus With lateral temporal lobes, medial temporal lobes, and precuneus This was demonstrated at the p<0001 level using a voxel-based multivariate analysis of covariance (MANCOVA) approach In addition, we used a data-driven technique of joint independent component analysis (ICA) that uncovers spatial pattern that are linked across modalities. It revealed a pattern of white matter tracts including cingulate bundle and associated fiber tracts resembling the findings from the hypothesis-driven analysis and was linked to the pattern of default mode network (DMN) connectivity in the resting state fMRI data Out findings support the notion that the functional connectivity between the posterior cingulate and hippocampus and the functional connectivity across the entire DMN is based oil distinct pattern of anatomical connectivity within the cerebral white matter (C) 2009 Elsevier Inc All rights reserved

Blind hyperspectral unmixing

This paper introduces a new hyperspectral unmixing method called Dependent Component Analysis (DECA). This method decomposes a hyperspectral image into a collection of reflectance (or radiance) spectra of the materials present in the scene (endmember signatures) and the corresponding abundance fractions at each pixel. DECA models the abundance fractions as mixtures of Dirichlet densities, thus enforcing the constraints on abundance fractions imposed by the acquisition process, namely non-negativity and constant sum. The mixing matrix is inferred by a generalized expectation-maximization (GEM) type algorithm. This method overcomes the limitations of unmixing methods based on Independent Component Analysis (ICA) and on geometrical based approaches. DECA performance is illustrated using simulated and real data.

Blind hyperspectral unmixing

Hyperspectral unmixing methods aim at the decomposition of a hyperspectral image into a collection endmember signatures, i.e., the radiance or reflectance of the materials present in the scene, and the correspondent abundance fractions at each pixel in the image. This paper introduces a new unmixing method termed dependent component analysis (DECA). This method is blind and fully automatic and it overcomes the limitations of unmixing methods based on Independent Component Analysis (ICA) and on geometrical based approaches. DECA is based on the linear mixture model, i.e., each pixel is a linear mixture of the endmembers signatures weighted by the correspondent abundance fractions. These abundances are modeled as mixtures of Dirichlet densities, thus enforcing the non-negativity and constant sum constraints, imposed by the acquisition process. The endmembers signatures are inferred by a generalized expectation-maximization (GEM) type algorithm. The paper illustrates the effectiveness of DECA on synthetic and real hyperspectral images.

Origin of Frontal Lobe Spikes in the Early Onset Benign Occipital Lobe Epilepsy (Panayiotopoulos Syndrome)

Objective: Early onset benign occipital lobe epilepsy (Panayiotopoulos syndrome [PS]) is a common and easily recognizable epilepsy. Interictal EEG spike activity is often multifocal but most frequently localized in the occipital lobes. The origin and clinical significance of the extra-occipital spikes remain poorly understood. Methods: Three patients with the PS and interictal EEG spikes with frontal lobe topography were studied using high-resolution EEG. Independent component analysis (ICA) was used to decompose the spikes in components with distinct temporal dynamics. The components were mapped in the scalp with a spline-laplacian algorithm. Results: The change in scalp potential topography from spike onset to peak, suggests the contribution of several intracranial generators, with different kinetics of activation and significant overlap. ICA was able to separate the major contributors to frontal spikes and consistently revealed an early activating group of components over the occipital areas in all the patients. The local origin of these early potentials was established by the spline-laplacian montage. Conclusions: Frontal spikes in PS are consistently associated with early and unilateral occipital lobe activation, suggesting a posteroanterior spike propagation. Significance: Frontal spikes in the PS represent a secondary activation triggered by occipital interictal discharges and do not represent an independent focus.

Processamento de imagens BOLD de Ressonância Magnética do músculo através da técnica ICA

Dissertação para obtenção do Grau de Mestre em Engenharia Biomédica

A new electrophysiological index for working memory load in humans.

Working memory, the ability to store and simultaneously manipulate information, is affected in several neuropsychiatric disorders which lead to severe cognitive and functional deficits. An electrophysiological marker for this process could help identify early cerebral function abnormalities. In subjects performing working memory-specific n-back tasks, event-related potential analysis revealed a positive-negative waveform (PNwm) component modulated in amplitude by working memory load. It occurs in the expected time range for this process, 140-280 ms after stimulus onset, superimposed on the classical P200 and N200 components. Independent Component Analysis extracted two functional components with latencies and topographical scalp distributions similar to the PNwm. Our results imply that the PNwm represents a new electrophysiological index for working memory load in humans.

Attention-related potentials allow for a highly accurate discrimination of mild cognitive impairment subtypes.

The three most frequent forms of mild cognitive impairment (MCI) are single-domain amnestic MCI (sd-aMCI), single-domain dysexecutive MCI (sd-dMCI) and multiple-domain amnestic MCI (md-aMCI). Brain imaging differences among single domain subgroups of MCI were recently reported supporting the idea that electroencephalography (EEG) functional hallmarks can be used to differentiate these subgroups. We performed event-related potential (ERP) measures and independent component analysis in 18 sd-aMCI, 13 sd-dMCI and 35 md-aMCI cases during the successful performance of the Attentional Network Test. Sensitivity and specificity analyses of ERP for the discrimination of MCI subgroups were also made. In center-cue and spatial-cue warning stimuli, contingent negative variation (CNV) was elicited in all MCI subgroups. Two independent components (ICA1 and 2) were superimposed in the time range on the CNV. The ICA2 was strongly reduced in sd-dMCI compared to sd-aMCI and md-aMCI (4.3 vs. 7.5% and 10.9% of the CNV component). The parietal P300 ERP latency increased significantly in sd-dMCI compared to md-aMCI and sd-aMCI for both congruent and incongruent conditions. This latency for incongruent targets allowed for a highly accurate separation of sd-dMCI from both sd-aMCI and md-aMCI with correct classification rates of 90 and 81%, respectively. This EEG parameter alone performed much better than neuropsychological testing in distinguishing sd-dMCI from md-aMCI. Our data reveal qualitative changes in the composition of the neural generators of CNV in sd-dMCI. In addition, they document an increased latency of the executive P300 component that may represent a highly accurate hallmark for the discrimination of this MCI subgroup in routine clinical settings.

Investigation of high-resolution functional magnetic resonance imaging by means of surface and array radiofrequency coils at 7 T.

In this investigation, high-resolution, 1x1x1-mm(3) functional magnetic resonance imaging (fMRI) at 7 T is performed using a multichannel array head coil and a surface coil approach. Scan geometry was optimized for each coil separately to exploit the strengths of both coils. Acquisitions with the surface coil focused on partial brain coverage, while whole-brain coverage fMRI experiments were performed with the array head coil. BOLD sensitivity in the occipital lobe was found to be higher with the surface coil than with the head array, suggesting that restriction of signal detection to the area of interest may be beneficial for localized activation studies. Performing independent component analysis (ICA) decomposition of the fMRI data, we consistently detected BOLD signal changes and resting state networks. In the surface coil data, a small negative BOLD response could be detected in these resting state network areas. Also in the data acquired with the surface coil, two distinct components of the positive BOLD signal were consistently observed. These two components were tentatively assigned to tissue and venous signal changes.

Comparing ICA-based and single-trial topographic ERP analyses.

Single-trial analysis of human electroencephalography (EEG) has been recently proposed for better understanding the contribution of individual subjects to a group-analysis effect as well as for investigating single-subject mechanisms. Independent Component Analysis (ICA) has been repeatedly applied to concatenated single-trial responses and at a single-subject level in order to extract those components that resemble activities of interest. More recently we have proposed a single-trial method based on topographic maps that determines which voltage configurations are reliably observed at the event-related potential (ERP) level taking advantage of repetitions across trials. Here, we investigated the correspondence between the maps obtained by ICA versus the topographies that we obtained by the single-trial clustering algorithm that best explained the variance of the ERP. To do this, we used exemplar data provided from the EEGLAB website that are based on a dataset from a visual target detection task. We show there to be robust correspondence both at the level of the activation time courses and at the level of voltage configurations of a subset of relevant maps. We additionally show the estimated inverse solution (based on low-resolution electromagnetic tomography) of two corresponding maps occurring at approximately 300 ms post-stimulus onset, as estimated by the two aforementioned approaches. The spatial distribution of the estimated sources significantly correlated and had in common a right parietal activation within Brodmann's Area (BA) 40. Despite their differences in terms of theoretical bases, the consistency between the results of these two approaches shows that their underlying assumptions are indeed compatible.

EEG windowed statistical wavelet scoring for evaluation and discrimination of muscular artifacts

EEG recordings are usually corrupted by spurious extra-cerebral artifacts, which should be rejected or cleaned up by the practitioner. Since manual screening of human EEGs is inherently error prone and might induce experimental bias, automatic artifact detection is an issue of importance. Automatic artifact detection is the best guarantee for objective and clean results. We present a new approach, based on the time–frequency shape of muscular artifacts, to achieve reliable and automatic scoring. The impact of muscular activity on the signal can be evaluated using this methodology by placing emphasis on the analysis of EEG activity. The method is used to discriminate evoked potentials from several types of recorded muscular artifacts—with a sensitivity of 98.8% and a specificity of 92.2%. Automatic cleaning ofEEGdata are then successfully realized using this method, combined with independent component analysis. The outcome of the automatic cleaning is then compared with the Slepian multitaper spectrum based technique introduced by Delorme et al (2007 Neuroimage 34 1443–9).

A Canonical Genetic Algorithm for Blind Inversion of Linear Channels

It is well known the relationship between source separation and blind deconvolution: If a filtered version of an unknown i.i.d. signal is observed, temporal independence between samples can be used to retrieve the original signal, in the same manner as spatial independence is used for source separation. In this paper we propose the use of a Genetic Algorithm (GA) to blindly invert linear channels. The use of GA is justified in the case of small number of samples, where other gradient-like methods fails because of poor estimation of statistics.

Automatic Recognition of Leaves by Shape Detection Pre-Processing with Ica

In this work we present a simulation of a recognition process with perimeter characterization of a simple plant leaves as a unique discriminating parameter. Data coding allowing for independence of leaves size and orientation may penalize performance recognition for some varieties. Border description sequences are then used to characterize the leaves. Independent Component Analysis (ICA) is then applied in order to study which is the best number of components to be considered for the classification task, implemented by means of an Artificial Neural Network (ANN). Obtained results with ICA as a pre-processing tool are satisfactory, and compared with some references our system improves the recognition success up to 80.8% depending on the number of considered independent components.

Coherency and sharpness measures by using ICA algorithms. An investigation for Alzheimer’s disease discrimination

In this paper, we present a comprehensive study of different Independent Component Analysis (ICA) algorithms for the calculation of coherency and sharpness of electroencephalogram (EEG) signals, in order to investigate the possibility of early detection of Alzheimer’s disease (AD). We found that ICA algorithms can help in the artifact rejection and noise reduction, improving the discriminative property of features in high frequency bands (specially in high alpha and beta ranges). In addition to different ICA algorithms, the optimum number of selected components is investigated, in order to help decision processes for future works.

Investigation of ICA algorithms for feature extraction of EEG signals in discrimination of Alzheimer disease

In this paper we present a quantitative comparisons of different independent component analysis (ICA) algorithms in order to investigate their potential use in preprocessing (such as noise reduction and feature extraction) the electroencephalogram (EEG) data for early detection of Alzhemier disease (AD) or discrimination between AD (or mild cognitive impairment, MCI) and age-match control subjects.