Higher-order co-occurrences for exploratory point pattern analysis and decision tree clustering on spatial data

Analyzing geographical patterns by collocating events, objects or their attributes has a long history in surveillance and monitoring, and is particularly applied in environmental contexts, such as ecology or epidemiology. The identification of patterns or structures at some scales can be addressed using spatial statistics, particularly marked point processes methodologies. Classification and regression trees are also related to this goal of finding "patterns" by deducing the hierarchy of influence of variables on a dependent outcome. Such variable selection methods have been applied to spatial data, but, often without explicitly acknowledging the spatial dependence. Many methods routinely used in exploratory point pattern analysis are2nd-order statistics, used in a univariate context, though there is also a wide literature on modelling methods for multivariate point pattern processes. This paper proposes an exploratory approach for multivariate spatial data using higher-order statistics built from co-occurrences of events or marks given by the point processes. A spatial entropy measure, derived from these multinomial distributions of co-occurrences at a given order, constitutes the basis of the proposed exploratory methods. © 2010 Elsevier Ltd.

Spatial pattern analysis of beta-amyloid (A beta) deposits in Alzheimer disease by linear regression

The spatial patterns of discrete beta-amyloid (Abeta) deposits in brain tissue from patients with Alzheimer disease (AD) were studied using a statistical method based on linear regression, the results being compared with the more conventional variance/mean (V/M) method. Both methods suggested that Abeta deposits occurred in clusters (400 to <12,800 mu m in diameter) in all but 1 of the 42 tissues examined. In many tissues, a regular periodicity of the Abeta deposit clusters parallel to the tissue boundary was observed. In 23 of 42 (55%) tissues, the two methods revealed essentially the same spatial patterns of Abeta deposits; in 15 of 42 (36%), the regression method indicated the presence of clusters at a scale not revealed by the V/M method; and in 4 of 42 (9%), there was no agreement between the two methods. Perceived advantages of the regression method are that there is a greater probability of detecting clustering at multiple scales, the dimension of larger Abeta clusters can be estimated more accurately, and the spacing between the clusters may be estimated. However, both methods may be useful, with the regression method providing greater resolution and the V/M method providing greater simplicity and ease of interpretation. Estimates of the distance between regularly spaced Abeta clusters were in the range 2,200-11,800 mu m, depending on tissue and cluster size. The regular periodicity of Abeta deposit clusters in many tissues would be consistent with their development in relation to clusters of neurons that give rise to specific neuronal projections.

The usefulness of spatial pattern analysis in understanding the pathogenesis of neurodegenerative disorders, with particular reference to plaque formation in Alzheimer's disease

Discrete, microscopic lesions are developed in the brain in a number of neurodegenerative diseases. These lesions may not be randomly distributed in the tissue but exhibit a spatial pattern, i.e., a departure from randomness towards regularlity or clustering. The spatial pattern of a lesion may reflect its development in relation to other brain lesions or to neuroanatomical structures. Hence, a study of spatial pattern may help to elucidate the pathogenesis of a lesion. A number of statistical methods can be used to study the spatial patterns of brain lesions. They range from simple tests of whether the distribution of a lesion departs from random to more complex methods which can detect clustering and the size, distribution and spacing of clusters. This paper reviews the uses and limitations of these methods as applied to neurodegenerative disorders, and in particular to senile plaque formation in Alzheimer's disease.

Higher-order co-occurrences for exploratory point pattern analysis and decision tree clustering on spatial data

Analyzing geographical patterns by collocating events, objects or their attributes has a long history in surveillance and monitoring, and is particularly applied in environmental contexts, such as ecology or epidemiology. The identification of patterns or structures at some scales can be addressed using spatial statistics, particularly marked point processes methodologies. Classification and regression trees are also related to this goal of finding "patterns" by deducing the hierarchy of influence of variables on a dependent outcome. Such variable selection methods have been applied to spatial data, but, often without explicitly acknowledging the spatial dependence. Many methods routinely used in exploratory point pattern analysis are2nd-order statistics, used in a univariate context, though there is also a wide literature on modelling methods for multivariate point pattern processes. This paper proposes an exploratory approach for multivariate spatial data using higher-order statistics built from co-occurrences of events or marks given by the point processes. A spatial entropy measure, derived from these multinomial distributions of co-occurrences at a given order, constitutes the basis of the proposed exploratory methods. © 2010 Elsevier Ltd.

A spatial pattern analysis of beta-amyloid (Abeta) deposition in the temporal lobe in Alzheimer's disease

To determine the factors influencing the distribution of -amyloid (Abeta) deposits in Alzheimer's disease (AD), the spatial patterns of the diffuse, primitive, and classic A deposits were studied from the superior temporal gyrus (STG) to sector CA4 of the hippocampus in six sporadic cases of the disease. In cortical gyri and in the CA sectors of the hippocampus, the Abeta deposits were distributed either in clusters 200-6400 microm in diameter that were regularly distributed parallel to the tissue boundary or in larger clusters greater than 6400 microm in diameter. In some regions, smaller clusters of Abeta deposits were aggregated into larger 'superclusters'. In many cortical gyri, the density of Abeta deposits was positively correlated with distance below the gyral crest. In the majority of regions, clusters of the diffuse, primitive, and classic deposits were not spatially correlated with each other. In two cases, double immunolabelled to reveal the Abeta deposits and blood vessels, the classic Abeta deposits were clustered around the larger diameter vessels. These results suggest a complex pattern of Abeta deposition in the temporal lobe in sporadic AD. A regular distribution of Abeta deposit clusters may reflect the degeneration of specific cortico-cortical and cortico-hippocampal pathways and the influence of the cerebral blood vessels. Large-scale clustering may reflect the aggregation of deposits in the depths of the sulci and the coalescence of smaller clusters.

A spatial pattern analysis of β-amyloid (Aβ) deposition in the temporal lobe in Alzheimer's disease

To determine the factors influencing the distribution of β-amyloid (Aβ) deposits in Alzheimer's disease (AD), the spatial patterns of the diffuse, primitive, and classic Aβ deposits were studied from the superior temporal gyrus (STG) to sector CA4 of the hippocampus in six sporadic cases of the disease. In cortical gyri and in the CA sectors of the hippocampus, the Aβ deposits were distributed either in clusters 200-6400 μm in diameter that were regularly distributed parallel to the tissue boundary or in larger clusters greater than 6400 μm in diameter. In some regions, smaller clusters of Aβ deposits were aggregated into larger 'superclusters'. In many cortical gyri, the density of Aβ deposits was positively correlated with distance below the gyral crest. In the majority of regions, clusters of the diffuse, primitive, and classic deposits were not spatially correlated with each other. In two cases, double immunolabelled to reveal the Aβ deposits and blood vessels, the classic Aβ deposits were clustered around the larger diameter vessels. These results suggest a complex pattern of Aβ deposition in the temporal lobe in sporadic AD. A regular distribution of Aβ deposit clusters may reflect the degeneration of specific cortico-cortical and cortico-hippocampal pathways and the influence of the cerebral blood vessels. Large-scale clustering may reflect the aggregation of deposits in the depths of the sulci and the coalescence of smaller clusters.

Density and spatial pattern of the neuropathological changes in the cerebellum in the prion disease variant Creutzfeldt-Jakob disease (vCJD)

The objective of this chapter is to quantify the neuropathology of the cerebellar cortex in cases of the prion disease variant Creutzfeldt-Jakob disease (vCJD). Hence, sequential sections of the cerebellum of 15 cases of vCJD were stained with H/E, or immunolabelled with a monoclonal antibody 12F10 against prion protein (PrP) and studied using quantitative techniques and spatial pattern analysis. A significant loss of Purkinje cells was evident in all cases. Densities of the vacuolation and the protease resistant form of prion protein (PrPSc) in the form of diffuse and florid plaques were greater in the granule cell layer (GL) than the molecular layer (ML). In the ML, vacuoles and PrPSc plaques, occurred in clusters which were regularly distributed along the folia, larger clusters of vacuoles and diffuse plaques being present in the GL. There was a negative spatial correlation between the vacuoles and the surviving Purkinje cells in the ML and a positive spatial correlation between the clusters of vacuoles and the diffuse PrPSc plaques in the ML and GL in five and six cases respectively. A canonical variate analysis (CVA) suggested a negative correlation between the densities of the vacuolation in the GL and the diffuse PrPSc plaques in the ML. The data suggest: 1) all laminae of the cerebellar cortex were affected by the pathology of vCJD, the GL more severely than the ML, 2) the pathology was topographically distributed especially in the Purkinje cell layer and GL, 3) pathological spread may occur in relation to a loop of anatomical projections connecting the cerebellum, thalamus, cerebral cortex, and pons, and 4) there are differences in the pathology of the cerebellum in vCJD compared with the M/M1 subtype of sporadic CJD (sCJD).

Using generative models for handwritten digit recognition

We describe a method of recognizing handwritten digits by fitting generative models that are built from deformable B-splines with Gaussian ``ink generators'' spaced along the length of the spline. The splines are adjusted using a novel elastic matching procedure based on the Expectation Maximization (EM) algorithm that maximizes the likelihood of the model generating the data. This approach has many advantages. (1) After identifying the model most likely to have generated the data, the system not only produces a classification of the digit but also a rich description of the instantiation parameters which can yield information such as the writing style. (2) During the process of explaining the image, generative models can perform recognition driven segmentation. (3) The method involves a relatively small number of parameters and hence training is relatively easy and fast. (4) Unlike many other recognition schemes it does not rely on some form of pre-normalization of input images, but can handle arbitrary scalings, translations and a limited degree of image rotation. We have demonstrated our method of fitting models to images does not get trapped in poor local minima. The main disadvantage of the method is it requires much more computation than more standard OCR techniques.

A hierarchical latent variable model for data visualization

Visualization has proven to be a powerful and widely-applicable tool the analysis and interpretation of data. Most visualization algorithms aim to find a projection from the data space down to a two-dimensional visualization space. However, for complex data sets living in a high-dimensional space it is unlikely that a single two-dimensional projection can reveal all of the interesting structure. We therefore introduce a hierarchical visualization algorithm which allows the complete data set to be visualized at the top level, with clusters and sub-clusters of data points visualized at deeper levels. The algorithm is based on a hierarchical mixture of latent variable models, whose parameters are estimated using the expectation-maximization algorithm. We demonstrate the principle of the approach first on a toy data set, and then apply the algorithm to the visualization of a synthetic data set in 12 dimensions obtained from a simulation of multi-phase flows in oil pipelines and to data in 36 dimensions derived from satellite images.

Hierarchical GTM: constructing localized non-linear projection manifolds in a principled way

It has been argued that a single two-dimensional visualization plot may not be sufficient to capture all of the interesting aspects of complex data sets, and therefore a hierarchical visualization system is desirable. In this paper we extend an existing locally linear hierarchical visualization system PhiVis ¸iteBishop98a in several directions: bf(1) We allow for em non-linear projection manifolds. The basic building block is the Generative Topographic Mapping (GTM). bf(2) We introduce a general formulation of hierarchical probabilistic models consisting of local probabilistic models organized in a hierarchical tree. General training equations are derived, regardless of the position of the model in the tree. bf(3) Using tools from differential geometry we derive expressions for local directional curvatures of the projection manifold. Like PhiVis, our system is statistically principled and is built interactively in a top-down fashion using the EM algorithm. It enables the user to interactively highlight those data in the ancestor visualization plots which are captured by a child model. We also incorporate into our system a hierarchical, locally selective representation of magnification factors and directional curvatures of the projection manifolds. Such information is important for further refinement of the hierarchical visualization plot, as well as for controlling the amount of regularization imposed on the local models. We demonstrate the principle of the approach on a toy data set and apply our system to two more complex 12- and 18-dimensional data sets.

What does the study of the spatial patterns of pathological lesions tell us about the pathogenesis of neurodegenerative disorders?

Discrete pathological lesions, which include extracellular protein deposits, intracellular inclusions and changes in cell morphology, occur in the brain in the majority of neurodegenerative disorders. These lesions are not randomly distributed in the brain but exhibit a spatial pattern, that is, a departure from randomness towards regularity or clustering. The spatial pattern of a lesion may reflect pathological processes affecting particular neuroanatomical structures and, therefore, studies of spatial pattern may help to elucidate the pathogenesis of a lesion and of the disorders themselves. The present article reviews first, the statistical methods used to detect spatial patterns and second, the types of spatial patterns exhibited by pathological lesions in a variety of disorders which include Alzheimer's disease, Down syndrome, dementia with Lewy bodies, Creutzfeldt-Jakob disease, Pick's disease and corticobasal degeneration. These studies suggest that despite the morphological and molecular diversity of brain lesions, they often exhibit a common type of spatial pattern (i.e. aggregation into clusters that are regularly distributed in the tissue). The pathogenic implications of spatial pattern analysis are discussed with reference to the individual disorders and to studies of neurodegeneration as a whole.

Hippocampal pathology in progressive supranuclear palsy (PSP): a quantitative study of 8 cases

Objective: To quantify the neuronal and glial cell pathology in the hippocampus and the parahippocampal gyrus (PHG) of 8 cases of progressive supranuclear palsy (PSP). Material: tau-immunolabeled sections of the temporal lobe of 8 diagnosed cases of PSP. Method: The densities of lesions were measured in the PHG, CA sectors of the hippocampus and the dentate gyrus (DG) and studied using spatial pattern analysis. Results: Neurofibrillary tangles (NFT) and abnormally enlarged neurons (EN) were most frequent in the PHG and in sector CA1 of the hippocampus, oligodendroglial inclusions (“coiled bodies”) (GI) in the PHG, subiculum, sectors CA1 and CA2, and neuritic plaques (NP) in sectors CA2 and CA4. The DG was the least affected region. Vacuolation and GI were observed in the alveus. No tufted astrocytes (TA) were observed. Pathological changes exhibited clustering, the lesions often exhibiting a regular distribution of the clusters parallel to the tissue boundary. There was a positive correlation between the degree of vacuolation in the alveus and the densities of NFT in CA1 and GI in CA1 and CA2. Conclusion: The pathology most significantly affected the output pathways of the hippocampus, lesions were topographically distributed, and hippocampal pathology may be one factor contributing to cognitive decline in PSP.

A quantitative study of the pathological changes in the cerebellum in 15 cases of variant Creutzfeldt–Jakob disease (vCJD)

Aims: To determine in the cerebellum in variant Creutzfeldt–Jakob disease (vCJD): (i) whether the pathology affected all laminae; (ii) the spatial topography of the pathology along the folia; (iii) spatial correlations between the pathological changes; and (iv) whether the pathology was similar to that of the common methionine/methionine Type 1 subtype of sporadic CJD. Methods: Sequential cerebellar sections of 15 cases of vCJD were stained with haematoxylin and eosin, or immunolabelled with monoclonal antibody 12F10 against prion protein (PrP) and studied using spatial pattern analysis. Results: Loss of Purkinje cells was evident compared with control cases. Densities of the vacuolation and the protease-resistant form of prion protein (PrPSc) (diffuse and florid plaques) were greater in the granule cell layer (GL) than the molecular layer (ML). In the ML, vacuoles and PrPSc plaques occurred in clusters regularly distributed along the folia with larger clusters of vacuoles and diffuse plaques in the GL. There was a negative spatial correlation between the vacuoles and the surviving Purkinje cells in the ML. There was a positive spatial correlation between the vacuoles and diffuse PrPSc plaques in the ML and GL. Conclusions: (i) all laminae were affected by the pathology, the GL more severely than the ML; (ii) the pathology was topographically distributed along the folia especially in the Purkinje cell layer and ML; (iii) pathological spread may occur in relation to the loop of anatomical connections involving the cerebellum, thalamus, cerebral cortex and pons; and (iv) there were pathological differences compared with methionine/methionine Type 1 sporadic CJD.

Spatial topography of the neurofibrillary tangles in cortical and subcortical regions in progressive supranuclear palsy

Objective: To study the topography of neurofibrillary tangles (NFT) in cortical and subcortical areas in progressive supranuclear palsy (PSP). Methods: Pattern analysis was carried out on tau-positive NFT in eight PSP cases. Results: Of the areas studied, NFT were randomly distributed in 68%, regularly distributed in 3%, and clustered in 29%. A regular distribution of clusters was more frequent in cortical than subcortical areas. Conclusion: NFT topography in subcortical areas was similar to inclusions in the synucleinopathy multiple system atrophy (MSA) but in cortical areas was comparable to other tauopathies. © 2006 Elsevier Ltd. All rights reserved.