Search

Mavigator is a graphical querying and browsing tool dedicated to naive users (computer non-professionals). It allows them to retrieve and analyse information from various data sources, in particular, from object-oriented and XML-oriented databases, providing a corresponding wrapper is implemented. Mavigator key concepts related to information retrieval include: intensional navigation, extensional navigation, and persistent baskets for recording temporary and final results. A basic problem related to end-user visual interfaces concerns final presentation and further processing of retrieval results. The Mavigator Active Extensions (AE) module allows the programmer to extend ad hoc the existing core functionalities through a program written in C#. Thus the retrieved data can be presented and analysed in any conceivable visual form. Another novel feature of Mavigator is Virtual Schemas, which make it possible to customize database schema, in particular, changing some names, adding virtual associations or hiding some classes. Virtual Schemas allow creating a customized version of an existing database schema and navigate within the database according to this schema. In this paper we present results of usability tests, which have been conducted within three independent student groups. The results of the tests suggest improvements to the prototype. [ABSTRACT FROM AUTHOR]

Experiences with e-books show that the principle obstacle to electronic reading is neither the weight, nor the autonomy or the discomfort of reading on screen, but the absence of reference mark which makes it possible to replace the current window of reading in the whole of the book. We present an automatic cartography of electronic documents which constitutes an attempt to facilitate navigation, reading and memorization of contents. It is to automatically build a singular picture which is designed to be associated to our remembering of each document, i.e. to our mental image. This picture corresponds to the cartography of an island. Its shape is build from the document structure; its coloring reflects the affective content of the text extracted by keywords spotting techniques while icons associated to document topics are added to textual legends as in ancient geographical maps. [ABSTRACT FROM AUTHOR]

This paper was designed to build an artificial multilayer network with the purpose of studying abilities like instant pattern recognition and discrimination where no learning would be required. The relevance refers to: (1) theories about putative biological mechanisms that would support innate perception, (2) technological implementation of faster systems for detection and classification of environmental stimulus without learning. Our model was built using few paradigmatic principles of neural organization. The connections obey a Gaussian function. When the network is submitted to diverse input patterns it produces both discriminative and distributed codes in all layers. Contrasting stimulus leads to an attention-like process by salience detection. Finally, the codes always hold a half of all nodes. [ABSTRACT FROM AUTHOR]

Brain functioning comprehension is an actual challenge. However the cerebral recordings produce a huge amount of data, it is difficult to select the pertinent data with the studied task. As we favor the dialog with the expert, we developed a methodology based on visualization. All steps are discussed and validated by the neurophysiologist as they still keep the link with the biological process. The presented method is based first on extraction of oscillatory information from time-frequency transform. These ones are organized in a graph structure. Finally graph-matching techniques bring indication signals variations such as frequency, time or power variations. [ABSTRACT FROM AUTHOR]

Morbidity and mortality are one of the most frequently used of statistics. It is hard to oversee the whole distribution of cases because the diagnoses spread on too many International Classification of Diseases codes (ICD 10th version). Usually, in practice, hospital managers are satisfied to study some pre-determined and/or ordered groups of data. ICDview will help to find these groups. The distribution of diagnoses is varying by time, location and several parameters. The ICD codes themselves have a main structure by the organ and kind of disease. Examining the cases by this technique can also show, which ICD classes must be regrouped. Caseview method bases on the Diagnosis Related Groups (DRGs). Somehundred pixels are pictured. The number of elements of the ICD tops more than ten-thousand. Because of this it is problematic to pixelize it, because it is very hard to picture such a big amount of data on one screen. ICDview uses the same like reference set, as the Caseview does. The ICD main groups are classified in the Caseview's columns. The groups of these main groups are pictured in these columns, but their order from the midline separating the medical and surgical entities can be determined on several ways. One way is to calculate it by the average prognosticated weight number of the cases at DRG. It can be calculated by the average prognosticated length of stay of the patients in hospital, and other ways too. The cases can be represented by the main diagnosis, the basic diagnosis or any other type of diagnosis, which stays in the background of the current cure. Examples of use of the method are given. [ABSTRACT FROM AUTHOR]

We describe a method turning scatterplot matrix visualizations into malleable graphical objects facilitating interaction and selection of pixelized data elements. The method relies on density estimation techniques [1,2] applied through standard image processing. A 2D scatterplot is considered as an image and is then transformed into nested regions that can be easily selected. Based on Wattenberg and Fisher [3], and as confirmed by our experience, we believe users have a good intuition interpreting and interacting with these multiscale graphical objects. Bio-molecular data serves here as a case study for our methodology. The method was discussed and designed in collaboration with experts in metabonomics and has proven to be useful and complementary to classical statistical methods. [ABSTRACT FROM AUTHOR]

Provided with evolved functionalities, a digestive endoscopy atlas can be used as a tool of training and even of diagnosis aid. The architecture of such a system is leaning on two bases, one of endoscopic knowledge another of case iconography. Being inspired by medical practice, a bi-leveled - disease knowledge base allows a classification of possible diagnoses and a case selection of the endoscopic case base, enabling the similarity step to complete the retrieval. This project benefits at many levels from the "pixelization paradigm". Indeed, to visualize the Knowledge and Case bases is of great interest, but it's more exciting to visualize the steps of the classification and of the similar case retrieval by the generation of images confronting the knowledge base and the case base to the new case. [ABSTRACT FROM AUTHOR]

A method for constructing condensed yet concise visualization of a long-term bioinformatics research project is presented. The chart combines information on data quality and chronology of research activities. It is implemented as Microsoft® Excel spreadsheet with VBA automation. The project comprised 3100 hours of EMG measurements in 23 pregnant sheep, each monitored for up to 5 months during pregnancy and immediately after labor. EMG signals were recorded from the surface of the uterine wall by electrodes implanted at the horn and cervix. The signals were assessed for quality, stored in electronic EMG archive and queried from there for visualization. The spreadsheet shows subject data as time-sorted column headings and pregnancy time along rows, with cell color depicting measurement efficacy rating (signal unusable, partly usable, OK), and a symbol marking labor date. The chart is best read if rotated into landscape format. Project group members confirmed usefulness of the chart in assessing research progress and spotting measurement problems. [ABSTRACT FROM AUTHOR]

This paper presents two tools developed initially to study neural networks information processing, but usable in any field that requires analyzing the behaviour of large data samples. One of the tools (visu) is devoted to the visualisation of simultaneously behaving events in several ways, either statically or dynamically along time, and allows examining each component individually. The other tool (xtms) is devoted to the analysis of temporal point process using a large number of different analysis, either involving single events or the interaction of two or three events, free running or cyclic. Examples in the domain of neural networks are given. [ABSTRACT FROM AUTHOR]

This paper proposes an approach to graphically represent the tracking data in Course Management Systems in order to mine and discover the usage of specific software modules. We have implemented a tool that aims to give, at a glance, a visual representation of the usage of a specific module in all the courses managed by the Course Management System. The tool also allows comparing the usage in different courses, and seeing the distribution of the usage over time. This information is useful to administrators of the Course Management System that have to know how much the modules have been used in courses. But also, it could be useful to instructional designers that have to design, plan, and evaluate the learning needs in institutions. [ABSTRACT FROM AUTHOR]

A huge amount of molecular data is available in protein data bank and various other libraries and this amount is increasing day by day. Devising new and efficient computational methods to extract useful information from this data is a big challenge for the researchers working in the field. Computational molecular docking refers to computational methods which attempt to obtain the best binding conformation of two interacting molecules. Information of the best binding conformation is useful in many applications such as rational drug design, recognition, cellular pathways, macromolecular assemblies, protein folding etc. Docking has three important aspects: (i) modeling of molecular shape, (ii) shape matching and (iii) scoring and ranking of potential solutions. In this paper, a new approach is proposed for shape matching in rigid body docking. The method gives visual information about the matching conformations of the molecules. In the approach proposed here, B-spline surface representation technique is used to model the patches of molecular surface. Surface normal and curvature properties are used to match these patches with each other. The 2-D approach used here for generation of surface patches is useful to pixellisation paradigm. [ABSTRACT FROM AUTHOR]

In surgical treatment of Parkinson's disease, deep brain stimulation requires high-precision positioning of electrodes, needing accurate localization and outlines of anatomical targets. Manual procedure of anatomical structures outlining on magnetic resonance images (MRI) takes about several hours. We proposed an automated localizing procedure aiming to shorten this task to some seconds. Different parameters were simultaneously assessed in our algorithm undertaking segmentation of anatomical structures. Intraclass correlation coefficients (ICCs) were computed for centers of gravity coordinates of structures between manual expert-mapped MRI and automated-mapped MRI. Tanimoto coefficients were computed accounting for pixels overlapping between these two procedures. Although ICCs showed almost perfect concordance, TC provided further information with a quite severe value about 35%. For both criteria, results were variable regarding each parameter in our process. With such complex results to relate, their presentations were enhanced using visualization methods resembling that of the generalized Case View method. [ABSTRACT FROM AUTHOR]

In most image fusion-based processes, image information are qualified by both numerical activities held by pixels or voxels (data domain) and spatial distribution of these values (spatial domain). Image data are often transformed (registration, multi-scale transform, etc.) early in the fusion process, thus losing a part of both their physical meaning and their numerical accuracy. We propose here a new image fusion scheme in which spatial information are managed apart from image activities, aiming at delaying the alteration of original data sets until the final aggregation/decision step of the process. The global idea is to independently model image information from the data and spatial domains, design fusion operators in both domains, and finally obtain the image aggregation model by combining these operators. Such a process makes it possible to introduce spatial coefficients resulting from spatial fusion into advanced aggregation models at the final step. The fusion in the spatial domain is based on discrete geometrical models of the images. It consists in applying a computational geometry algorithm stemming from the study of the classical digital coordinates changing problem, and modified to be efficient even on large 3D images. Two applications of the fusion process are proposed in the field of medical image analysis, for brain image synthesis and activity quantification, mainly destined to the automated diagnosis of Parkinsonian syndromes. [ABSTRACT FROM AUTHOR]

Visualization of the massive datasets needs new methods which are able to quickly and easily reveal their contents. The projection of the data cloud is an interesting paradigm in spite of its difficulty to be explored when data plots are too numerous. So we study a new way to show a bidimensional projection from a multidimensional data cloud: our generative model constructs a tabular view of the projected cloud. We are able to show the high densities areas by their non equidistributed discretization. This approach is an alternative to the self-organizing map when a projection does already exist. The resulting pixel views of a dataset are illustrated by projecting a data sample of real images: it becomes possible to observe how are laid out the class labels or the frequencies of a group of modalities without being lost because of a zoom enlarging change for instance. The conclusion gives perspectives to this original promising point of view to get a readable projection for a statistical data analysis of large data samples. [ABSTRACT FROM AUTHOR]

We propose in this paper a visualization approach for large online databases using the Hilbert space-filling curve to map N-dimensional data points to 2D or 3D points. Dimensionality reduction methods like principal component analysis (PCA), multi dimensional scaling (MDS) or self organizing maps (SOMS) can map N-dimensional data points with N>>3 into 3 dimensional or 2 dimensional values that allow us to visualize the data. These methods although popular, require either the calculation of a scatter matrix, eigenvalues and eigenvectors, or the iteration of learning algorithms. Therefore these methods cannot perform online, can be slow with large databases and always produce information loss when the data is mapped from the multidimensional space to the 2D or 3D image. Space-filling curves like the Peano, Z, and Hilbert curve, on the contrary, produce a 1-to-1 mapping between points in a line segment and an arbitrary N-Dimensional hypercube. This 1-to-1 mapping guarantees that there is no information loss on the transformation. Specifically the Hilbert space-filling curve is known to preserve the Lebesgue measure and has been proven to produce an optimal mapping in the sense that an arbitrary contiguous block of information will receive the minimum number of splits in the mapped space. The Hilbert space-filling curve has been extensively used for indexing and clustering by mapping N-dimensional data points to 1-dimensional values. We propose here to use the curve to map to 2 or 3 dimensions for purposes of visualization: By taking advantage of its 1-to-1 nature, a new and generic method to map N-dimensional data points to 2D or 3D points using the Hilbert space-filling curve is developed. We prove theoretically that the calculation of the mapping can be done in constant time if we fix the order of approximation, thereby giving linear O(n) performance on the number of data points to map. We create a Hilbert space-filling curve visualization tool that is much faster than the other methods mentioned and allows us to generate quickly for very large datasets various different visualizations of the data, thereby compensating the lack of use of statistical information in the calculation of the mapped points. We compare our approach to MDS and PCA with a benchmark data set and three real datasets using the distance preserving and topology preserving measure as benchmarks. Our experiments indicate that the Hilbert space-filling curve produces acceptable quality of mapping while achieving much faster visualization and is therefore especially useful for online visualization of very large data sets. [ABSTRACT FROM AUTHOR]

Pixelization is the simple yet powerful technique of mapping each element of some data set to a pixel in a 2D image. There are 2 primary characteristics of pixels that can be leveraged to impart information: 1. their color and color-related attributes (hue, saturation, etc.) and 2. their arrangement in the image. We have found that applying a dimensional stacking layout to pixelization uniquely facilitates feature discovery, informs and directs user queries, supports interactive data mining, and provides a means for exploratory analysis. In this paper we describe our approach and how it is being used to analyze multidimensional, multivariate neuroscience data. [ABSTRACT FROM AUTHOR]

Multidimensional databases are commonly used for decision making in the context of data warehouses. Considering the multidimensional model, data are presented as hypercubes organized according to several dimensions. However, in general, hypercubes have more than three dimensions and contain a huge amount of data, and so cannot be easily visualized. In this paper, we show that data cubes can be visualized as images by building blocks that contain mostly the same value. Blocks are built up using an APriori-like algorithm and each block is considered as a set of pixels which colors depend on the corresponding value. The key point of our approach is to set how to display a given block according to its corresponding value while taking into account that blocks may overlap. In this paper, we address this issue based on the Pixelization paradigm. [ABSTRACT FROM AUTHOR]

One of the key characteristics of the clinical practice of anatomic pathology is turnaround time. However, standard methods of reporting and data representation provide limited insight. Many factors have to be taken into consideration. Significance of various points on a timescale varies depending on days of the week and holidays. The number of pathologists, subspecialty, and case mix- all has to be in front of an observer to understand events and their circumstances. Currently, reporting requires printing of numerous tabular data sets that provide different views of the same event. Through the use of Java 2D API, we demonstrate how 2D graphic, color gradient and interactive display of data delivered via the Web could facilitate analysis of a particular problem of turnaround time in anatomic pathology operations and similar complex events. [ABSTRACT FROM AUTHOR]

A heat-map type of chart for depicting large number of cases and up to twenty-five categorical variables with spreadsheet software is presented. It is implemented in Microsoft® Excel using standard formulas, sorting and simple VBA code. The motivating example depicts accuracy of automated assignment of MeSH® descriptor headings to abstracts of medical articles. Within each abstract, predicted support for each heading is ranked, then for each heading actually assigned/non-assigned by human specialist (depicted by black/white cell), high/low support is depicted on nine-point two-colour scale. Thus, each case (abstract) is depicted by one row of a table and each variable (heading) with two adjacent columns. Rank-based classification accuracy measure is calculated for each case, and rows are sorted in increasing accuracy order downwards. Based on analogous measure, variables are sorted in increasing prediction accuracy order rightwards. Another biomedical dataset is presented with a similar chart. Different methods for predicting binary outcomes can be visualised, and the procedure is easily extended to polytomous variables. [ABSTRACT FROM AUTHOR]

Data mining involves useful knowledge discovery using a data matrix consisting of records and attributes or variables. Not all the attributes may be useful in knowledge discovery, as some of them may be redundant, irrelevant, noisy or even opposing. Furthermore, using all the attributes increases the complexity of solving the problem. The Minimum Attribute Subset Selection Problem (MASSP) has been studied for well over three decades and researchers have come up with several solutions In this paper a new technique is proposed for the MASSP based on the crossing minimization paradigm from the domain of graph drawing using biclustering. Biclustering is used to quickly identify those attributes that are significant in the data matrix. The attributes identified are then used to perform one-way clustering and generate pixelized visualization of the clustered results. Using the proposed technique on two real datasets has shown promising results. [ABSTRACT FROM AUTHOR]

We present new visual data mining algorithms for interactive decision tree construction with large datasets. The size of data stored in the world is constantly increasing but the limits of current visual data mining (and visualization) methods concerning the number of items and dimensions of the dataset treated are well known (even with pixellisation methods). One solution to improve these methods is to use a higher-level representation of the data, for example a symbolic data representation. Our new interactive decision tree construction algorithms deal with interval and taxonomical data. With such a representation, we are able to deal with potentially very large datasets because we do not use the original data but higher-level data representation. Interactive algorithms are examples of new data mining approach aiming at involving more intensively the user in the process. The main advantages of this user-centered approach are the increased confidence and comprehensibility of the obtained model, because the user was involved in its construction and the possible use of human pattern recognition capabilities. We present some results we obtained on very large datasets. [ABSTRACT FROM AUTHOR]

Pixel-based visualization techniques have proven to be of high value in visual data exploration, since mapping data points to pixels not only allows the analysis and visualization of large data sets, but also provides an intuitive way to convert raw data into a graphical form that often fosters new insights, encouraging the formation and validation of new hypotheses to the end of better problem solving and gaining deeper domain knowledge. But the ever increasing mass of information leads to new challenges on pixel-based techniques and concepts, since the volume, complexity and dynamic nature of today's scientific and commercial data sets are beyond the capability of many of current presentation techniques. Most existing pixel based approaches do not scale well on such large data sets as visual representation suffers from the high number of relevant data points, that might be even higher than the available monitor resolution and does therefore not allow a direct mapping of all data points to pixels on the display. In this paper we focuses on ways to increase the scalability of pixel based approaches by integrating relevance driven techniques into the visualization process. We provide first examples for effective scalable pixel based visualizations of financial- and geo-spatial data. [ABSTRACT FROM AUTHOR]

Various approaches exist related to Pixelization Paradigm. These approaches can be methodological or applied. Therefore finding formalism able to grasp all these approaches appears important. This formalism comes from category theory, the main formal tool being the functor notion. A category is characterized by objects and morphisms between these objects. Two categories can be associated thanks to a functor that respectively maps objects and morphisms from one category to objects and morphims of the other category. A pixelized visualization is defined by an object and two morphisms. The object is the reference frame of the visualization; the morphisms respectively associate a value and a colour to each cell of the reference frame. Functors formally allow implementing transformations on the visualization. The approach on the one handallowsidentifying various atomic notions of Pixelization paradigm with formal entities and on the other hand, thanks to this, allows studying them and opening new ways. Various illustrations are given. [ABSTRACT FROM AUTHOR]