John Hiller

Rethinking online monitoring methods for information retrieval systems: from search product to search process

Searching information retrieval systems is a highly interactive, iterative process that cannot be understood simply by comparing the output of a search session (the “search product”) to a query stated in advance. In this article, we examine evaluation goals and methods for studying information retrieval behavior, drawing examples from our own research and that of others. We limit our review to research that employs online monitoring, also known as transaction log analysis. Online monitoring is one of few methods that can capture detailed data on the search process at a reasonable cost; these data can be used to build quantitative models or to support qualitative interpretations of quantitative results. Monitoring is a data collection technique rather than a research design, and can be employed in experimental or field studies, whether alone or combined with other data collection methods. Based on the research questions of interest, the researcher must determine what variables to collect from each data source, which to treat as independent variables to manipulate, and which to treat as dependent variables to observe effects. Studies of searching behavior often treat search task and searcher characteristics as independent variables and may manipulate other independent variables specific to the research questions addressed. Search outcomes, time, and search paths frequently are treated as dependent variables. We discuss each of these sets of variables, illustrating them with sample results from the literature and from our own research. Our examples are drawn from the Science Library Catalog project, a 7-year study of childrens searching behavior on an experimental retrieval system. We close with a brief discussion of the implications of these results for the design of information retrieval systems.

TOTAL UNCERTAINTY REVISITED

A recently proposed [Pal et al., 1992] measure of uncertainty for evidence assignments is analyzed. Termed total uncertainty, it becomes, for assignment We review its axiomatic characterization and properties, establish its relationship with probabilistic entropy, and discuss its extremal properties. We then demonstrate that TU(m) is a representative instance of a class of related measures. which share the same axiomatic properties. Finally, we apply TU(m) to possibilistic assignments. Find its maximum, and compare to that of possibilistic information U( p).

Adaptive nonlinear diffusion algorithm for image filtering

The nonlinear anisotropic diffusive process has shown the good property of eliminating noise while preserving the accuracy of edges, and has been widely used in image processing. However, filtering depends on the threshold of the diffusion process, i.e., the cut-off contrast of edges. The threshold varies form image to image and even from region to region within an image. The problem compounds with intensity distortion and contrast variation. We have developed an adaptive diffusion scheme by applying the Central Limit Theorem to selecting the threshold. Gaussian distribution and Rayleigh distribution are used to estimate the distributions of visual objects in images. Regression under such distributions separates the distribution of the major object from other visual objects in a single peak histogram. The separation helps to automatically determine the threshold. A fast algorithm is derived for the regression process. The method has been successfully used in filtering various medical images.

Modelling the real world by Multi-World data model

This paper discusses various philosophical issues that relate to database modelling. In particular, we argue that existing paradigms fail in a multi-database environment, and hence it needs to be supplanted by one that is more flexible and dynamic. Based on this new paradigm we propose a MultiWorld data model which consists of clusters of private/public databases integrated via a contracting mechanism. Each database is autonomous but is able to retrieve external as well as local data. Autonomy appears to lead to conflict. We discuss its nature and propose a resolution strategy based on combinatory theories of truth.<<ETX>>

A Unifying Model of Data, Metadata and Context

In this paper, we define Context as a data structure that gives the meaning and the environmental information of the detail data, and discuss two of its main features-nested structure and role dynamism. Then we propose a Context Model that unifies detail, meta and contextual data. We also discuss a Context-oriented query mechanism and illustrate the model by query examples.

Development of medical image databases

Current use of medical imaging is far from satisfactory in terms of cost/effectiveness. To improve this, it is necessary to use database techniques. Different from a digital archive, a medical image database is capable of information abstraction, inferencing and reasoning. We will review current research and techniques world-wide, especially, the KMeD system developed in UCLA. We then address various issues in the development of medical image databases and present our work in the field. Many issues are also general concerns in medical imaging such as filtering, segmentation, registration, clustering, compression, reconstruction and visualization.

Anatomic vascular phantom for the verification of MRA and XRA visualization and fusion

A project is underway to develop automated methods of fusing cerebral magnetic resonance angiography (MRA) and x-ray angiography (XRA) for creating accurate visualizations used in planning treatment of vascular disease. We have developed a vascular phantom suitable for testing segmentation and fusion algorithms with either derived images (psuedo-MRA/psuedo-XRA) or actual MRA or XRA image sequences. The initial unilateral arterial phantom design, based on normal human anatomy, contains 48 tapering vascular segments with lumen diameters from 2.5 millimeter to 0.25 millimeter. The initial phantom used rapid prototyping technology (stereolithography) with a 0.9 millimeter vessel wall fabricated in an ultraviolet-cured plastic. The model fabrication resulted in a hollow vessel model comprising the internal carotid artery, the ophthalmic artery, and the proximal segments of the anterior, middle, and posterior cerebral arteries. The complete model was fabricated but the models lumen could not be cleared for vessels with less than 1 millimeter diameter. Measurements of selected vascular outer diameters as judged against the CAD specification showed an accuracy of 0.14 mm and precision (standard deviation) of 0.15 mm. The plastic vascular model produced provides a fixed geometric framework for the evaluation of imaging protocols and the development of algorithms for both segmentation and fusion.

Visualizing cerebral vasculature: imaging limitations from 2D and 3D data

Arterio-venous malformations (AVMs) are a congenital disorder that affects a small percentage of the population. They are treated by blocking or reducing the blood supply followed by surgery. This paper looks in a preliminary way at visualizing the cerebral vasculature and ultimately the AVMs. These visualizations provide support for the surgeons and radiologists. Our concern is to substantiate the point that there are deficiencies in the data correctable with reference to digital subtraction angiograms and we conjecture that knowledge based processing of this data may lead to improved results. The paper explores the basis of the difficulty and it compares the performance of several algorithms. Simple geometric objects are studied and the dependence of error on several parameters is shown. A comparison is drawn between the richness of the data available from x-ray angiograms (XRAs) and magnetic resonance angiograms (MRAs). Inferences are drawn on approaches that may be appropriate for the evolution of a description of the vasculature. Comment is also made on the way in which different representations may be compared.

High resolution anatomic model fabrication from CT: Visualizing 3-D error

A new CAM rapid prototype technology is being used to produce life-size anatomic models for custom prosthesis design and surgical planning. An 80 slice CT study was used to produce three bone models with stereolithography, each employing variations in solid models derived from segmented CT data. The error in linear measurements between the CT data and the fabricated model is discussed and a procedure involving a secondary CT scan of the plastic fabricated model and subsequent registration and visualization of error is presented.

Automatic segmentation applied to obstetric images

This paper presents a shape-based approach for searching and extracting fetal skull boundaries from an obstetric image. The proposed method relies on two major steps. Firstly, we apply the reference axes to scan the image for all potential skull boundaries. The possible skull boundaries are determined whether they are candidates. The candidate with the highest confident value will be selected as the expected head boundary. Then, the position of the expected head boundary is initialized. Secondly, we refine the initial skull boundary using the fuzzy contour model modified from the active contour basis. This results the continuous and smooth fetal skull boundary that we can use for the medical parameter measurement.