Analysis of brain activation patterns using a 3-D scale-space primal sketch

Impact of scale space search on age- and gender-related changes in MRI-based cortical morphometry

In magnetic resonance imaging based brain morphometry, Gaussian smoothing is often applied to increase the signal-to-noise ratio and to increase the detection power of statistical parametric maps. However, most existing studies used a single smoothing filter without adequately justifying their choices. In this article, we want to determine the extent for which performing a morphometry analysis using multiple smoothing filters, namely conducting a scale space search, improves or decreases the detection power. We first compared scale space search with single-filter analysis through a simulated population study. The multiple comparisons in our four-dimensional scale space searches were corrected for using a unified P-value approach. Our results illustrate that, compared with a single-filter analysis, a scale space search analysis can properly capture the variations in analysis results caused by variations in smoothing, and more importantly, it can obviously increase the sensitivity for detecting brain morphometric changes. We also show that the cost of an increased critical threshold for conducting a scale space search is very small. In the second experiment, we investigated age and gender effects on cortical volume, thickness, and surface area in 104 normal subjects using scale space search. The obtained results provide a perspective of scale space theory on the morphological changes with age and gender. These results suggest that, in exploratory studies of aging, gender, and disease, conducting a scale space search is essential, if we are to produce a complete description of the structural changes or abnormalities associated with these dimensions.

A topographic latent source model for fMRI data

We describe and evaluate a new statistical generative model of functional magnetic resonance imaging (fMRI) data. The model, topographic latent source analysis (TLSA), assumes that fMRI images are generated by a covariate-dependent superposition of latent sources. These sources are defined in terms of basis functions over space. The number of parameters in the model does not depend on the number of voxels, enabling a parsimonious description of activity patterns that avoids many of the pitfalls of traditional voxel-based approaches. We develop a multi-subject extension where latent sources at the subject-level are perturbations of a group-level template. We evaluate TLSA according to prediction, reconstruction and reproducibility. We show that it compares favorably to a Naive Bayes model while using fewer parameters. We also describe a hypothesis testing framework that can be used to identify significant latent sources.

Variability of fMRI-response patterns at different spatial observation scales

Functional organization units of the cerebral cortex exist over a wide range of spatial scales, from local circuits to entire cortical areas. In the last decades, scale-space representations of neuroimaging data suited to probe the multi-scale nature of cortical functional organization have been introduced and methodologically elaborated. For this purpose, responses are statistically detected over a range of spatial scales using a family of Gaussian filters, with small filters being related to fine and large filters-to coarse spatial scales. The goal of the present study was to investigate the degree of variability of fMRI-response patterns over a broad range of observation scales. To this aim, the same fMRI data set obtained from 18 subjects during a visuomotor task was analyzed with a range of filters from 4- to 16-mm full width at half-maximum (FWHM). We found substantial observation-scale-related variability. For example, using filter widths of 6- to 8-mm FWHM, in the group-level results, significant responses in the right secondary visual but not in the primary visual cortex were detected. However, when larger filters were used, the responses in the right primary visual cortex reached significance. Often, responses in probabilistically defined areas were significant when both small and large filters, but not intermediate filter widths were applied. This suggests that brain responses can be organized in local clusters of multiple distinct activation foci. Our findings illustrate the potential of multi-scale fMRI analysis to reveal novel features in the spatial organization of human brain responses.

Feature-based morphometry: discovering group-related anatomical patterns

This paper presents feature-based morphometry (FBM), a new fully data-driven technique for discovering patterns of group-related anatomical structure in volumetric imagery. In contrast to most morphometry methods which assume one-to-one correspondence between subjects, FBM explicitly aims to identify distinctive anatomical patterns that may only be present in subsets of subjects, due to disease or anatomical variability. The image is modeled as a collage of generic, localized image features that need not be present in all subjects. Scale-space theory is applied to analyze image features at the characteristic scale of underlying anatomical structures, instead of at arbitrary scales such as global or voxel-level. A probabilistic model describes features in terms of their appearance, geometry, and relationship to subject groups, and is automatically learned from a set of subject images and group labels. Features resulting from learning correspond to group-related anatomical structures that can potentially be used as image biomarkers of disease or as a basis for computer-aided diagnosis. The relationship between features and groups is quantified by the likelihood of feature occurrence within a specific group vs. the rest of the population, and feature significance is quantified in terms of the false discovery rate. Experiments validate FBM clinically in the analysis of normal (NC) and Alzheimer's (AD) brain images using the freely available OASIS database. FBM automatically identifies known structural differences between NC and AD subjects in a fully data-driven fashion, and an equal error classification rate of 0.80 is achieved for subjects aged 60-80 years exhibiting mild AD (CDR=1).

Coordinate-based versus structural approaches to brain image analysis

A basic issue in neurosciences is to look for possible relationships between brain architecture and cognitive models. The lack of architectural information in magnetic resonance images, however, has led the neuroimaging community to develop brain mapping strategies based on various coordinate systems without accurate architectural content. Therefore, the relationships between architectural and functional brain organizations are difficult to study when analyzing neuroimaging experiments. This paper advocates that the design of new brain image analysis methods inspired by the structural strategies often used in computer vision may provide better ways to address these relationships. The key point underlying this new framework is the conversion of the raw images into structural representations before analysis. These representations are made up of data-driven elementary features like activated clusters, cortical folds or fiber bundles. Two classes of methods are introduced. Inference of structural models via matching across a set of individuals is described first. This inference problem is illustrated by the group analysis of functional statistical parametric maps (SPMs). Then, the matching of new individual data with a priori known structural models is described, using the recognition of the cortical sulci as a prototypical example.

Differential metabolic activity in the striosome and matrix compartments of the rat striatum during natural behaviors

The striosome and matrix compartments of the striatum are clearly identified by their neurochemical expression patterns and anatomical connections. To determine whether these compartments are distinguishable functionally, we used [14C]deoxyglucose metabolic mapping in the rat and tested whether neutral behavioral states (free movement, gentle restraint, and focal tactile stimulation under gentle restraint) were associated with regions of high metabolic activity in the matrix, in striosomes, or in both. We identified metabolic peaks in the striatum by means of image analysis, striosome-matrix boundaries by [3H]naloxone binding, and primary somatosensory corticostriatal input clusters by injections of anterograde tracer into electrophysiologically identified sites in SI. Peak metabolic activity was primarily confined to the matrix compartment under each behavioral condition. These findings show that during relatively neutral behavioral conditions the balance of activity between the two compartments favors the matrix and suggest that this balance is present in the striatum as part of normal behavior and processing of afferent activity.

A database generator for human brain imaging

Sharing scientific data containing complex information requires new concepts and new technology. NEUROGENERATOR is a database generator for the neuroimaging community. A database generator is a database that generates new databases. The scientists submit raw PET and fMRI data to NEUROGENERATOR, which then processes the data in a uniform way to create databases of homogeneous data suitable for data sharing, met-analysis and modelling the human brain at the systems level. These databases are then distributed to the scientists.

Evaluation of using absolute versus relative base level when analyzing brain activation images using the scale-space primal sketch

A dominant approach to brain mapping is to define functional regions in the brain by analyzing images of brain activation obtained from positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). This paper presents an evaluation of using one such tool, called the scale-space primal sketch, for brain activation analysis. A comparison is made concerning two possible definitions of a significance measure of blob structures in scale-space, where local contrast is measured either relative to a local or global reference level. Experiments on real brain data show that (i) the global approach with absolute base level has a higher degree of correspondence to a traditional statistical method than a local approach with relative base level, and that (ii) the global approach with absolute base level gives a higher significance to small blobs that are superimposed on larger scale structures, whereas the significance of isolated blobs largely remains unaffected. Relative to previously reported works, the following two technical improvements are also presented. (i) A post-processing tool is introduced for merging blobs that are multiple responses to image structures. This simplifies automated analysis from the scale-space primal sketch. (ii) A new approach is introduced for scale-space normalization of the significance measure, by collecting reference statistics of residual noise images obtained from the general linear model.

Data mining in brain imaging

Data mining in brain imaging is proving to be an effective methodology for disease prognosis and prevention. This, together with the rapid accumulation of massive heterogeneous data sets, motivates the need for efficient methods that filter, clarify, assess, correlate and cluster brain-related information. Here, we present data mining methods that have been or could be employed in the analysis of brain images. These methods address two types of brain imaging data: structural and functional. We introduce statistical methods that aid the discovery of interesting associations and patterns between brain images and other clinical data. We consider several applications of these methods, such as the analysis of task-activation, lesion-deficit, and structure morphological variability; the development of probabilistic atlases; and tumour analysis. We include examples of applications to real brain data. Several data mining issues, such as that of method validation or verification, are also discussed.

Structural group analysis of functional activation maps

We present here a new method for cerebral activation detection over a group of subjects. This method is performed using individual activation maps of any sort. It aims at processing a group analysis while preserving individual information and at overcoming as far as possible limitations of the spatial normalization used to compare different subjects. We designed it such that it provides the individual occurrence of the activations detected at a group level. The localization can then be performed on the individual anatomy of each subject. The analysis starts with a hierarchical multiscale object-based description of each individual map. These descriptions are then compared, rather than comparing the images directly. The analysis is thus performed at an object level instead of voxel by voxel. It is made using a comparison graph, on which a labeling process is performed. The label field on the graph is modeled by a Markov random field, which allows us to introduce high-level rules of interrogation of the data. The process has been evaluated on simulated data and real data from a PET protocol.