Bong-Soo Sohn

Segmentation of Regions of Interest in Mammograms in a Topographic Approach

This paper presents a novel method for the segmentation of regions of interest in mammograms. The algorithm concurrently delineates the boundaries of the breast boundary, the pectoral muscle, as well as dense regions that include candidate masses. The resulting representation constitutes an analysis of the global structure of the object in the mammogram. We propose a topographic representation called the isocontour map, in which a salient region forms a dense quasi-concentric pattern of contours. The topological and geometrical structure of the image is analyzed using an inclusion tree that is a hierarchical representation of the enclosure relationships between contours. The ¿saliency¿ of a region is measured topologically as the minimum nesting depth. Features at various scales are analyzed in multiscale isocontour maps, and we demonstrate that the multiscale scheme provides an efficient way of achieving better delineations. Experimental results demonstrate that the proposed method has potential as the basis for a prompting system in mammogram mass detection.

Time-varying contour topology

The contour tree has been used to compute the topology of isosurfaces, generate a minimal seed set for accelerated isosurface extraction, and provide a user interface to segment individual contour components in a scalar field. In this paper, we extend the benefits of the contour tree to time-varying data visualization. We define temporal correspondence of contour components and describe an algorithm to compute the correspondence information in time-dependent contour trees. A graph representing the topology changes of time-varying isosurfaces is constructed in real-time for any selected isovalue using the precomputed correspondence information. Quantitative properties, such as surface area and volume of contour components, are computed and labeled on the graph. This topology change graph helps users to detect significant topological and geometric changes in time-varying isosurfaces. The graph is also used as an interactive user interface to segment, track, and visualize the evolution of any selected contour components over time.

Adaptive and quality 3D meshing from imaging data

This paper presents an algorithm to extract adaptive and quality 3D meshes directly from volumetric imaging data - primarily Computed Tomography (CT) and Magnetic Resonance Imaging (MRI). The extracted tetrahedral and hexahedral meshes are extensively used in finite element simulations. Our comprehensive approach combines bilateral and anisotropic (feature specific) diffusion filtering, with contour spectrum based, isosurface and interval volume selection. Next, a top-down octree subdivision coupled with the dual contouring method is used to rapidly extract adaptive 3D finite element meshes from volumetric imaging data. The main contributions are extending the dual contouring method to crack free interval volume tetrahedralization and hexahedralization with feature sensitive adaptation. Compared to other tetrahedral extraction methods from imaging data, our method generates better quality adaptive 3D meshes without hanging nodes. Our method has the properties of crack prevention and feature sensitivity.

Feature based volumetric video compression for interactive playback

In this paper, we describe a compression scheme for encoding time-varying isosurfaces and amorphous volumetric features (volumes within specified value ranges) in a unified way, which allows for on-line reconstruction and rendering. Since the size of even one frame in a time-varying data set is very large, transmission and online reconstruction are the main bottlenecks for interactive visualization of time-varying volume and surface data. To increase the run-time decompression speed and compression ratio, we decompose the volume into small blocks and encode only the significant blocks that contribute to the isosurface and volumetric features. The result shows that our compression scheme achieves high compression ratio with fast reconstruction, which is effective for client-side rendering of time-varying isosurfaces with amorphous volumetric features.

Topology Preserving Tetrahedral Decomposition Applied To Trilinear Interval Volume Tetrahedrization

We describe a method to decompose a cube with trilinear interpolation into a collection of tetrahedra with linear interpolation, where the isosurface topology is preserved for all isovalues during decomposition. Visualization algorithms that require input scalar data to be defined on a tetrahedral grid can utilize our method to process 3D rectilinear data with topological correctness. As one of many possible examples, we apply the decomposition method to topologically accurate tetrahedral mesh extraction of an interval volume from trilinear volumetric imaging data. The topological correctness of the resulting mesh can be critical for accurate simulation and visualization.

Volumetric video compression for interactive playback

We develop a volumetric video system which supports interactive browsing of compressed time-varying volumetric features (significant isosurfaces and interval volumes). Since the size of even one volumetric frame in a time-varying 3D data set is very large, transmission and on-line reconstruction are the main bottlenecks for interactive remote visualization of time-varying volume and surface data. We describe a compression scheme for encoding time-varying volumetric features in a unified way, which allows for on-line reconstruction and rendering. To increase the run-time decompression speed and compression ratio, we decompose the volume into small blocks and encode only the significant blocks that contribute to the isosurfaces and interval volumes. The results show that our compression scheme achieves high compression ratio with fast reconstruction, which is effective for client-side rendering of time-varying volumetric features.

Topology Preserving Tetrahedral Decomposition of Trilinear Cell

We describe a method to decompose a cube with trilinear interpolation into a set of tetrahedra with linear interpolation, where isosurface topology is preserved during decomposition for all isovalues. This method is useful for converting from a rectilinear grid into a tetrahedral grid in scalar data with topological correctness. We apply our method to topologically and geometrically accurate isosurface extraction.

Bas-relief generation from face photograph based on facial feature enhancement

This paper describes a novel method for generating a bas-relief surface from the photographic image of a human face. One of the simplest methods is to take each pixel brightness as a depth value and use it to elevate the resulting surface. Although this approach can generate a bas-relief surface with realistic textures, it has the disadvantage of generating erroneous 3D depth. This problem is especially serious in the areas of facial features, such as hair, eyes, eyebrows, nose, and lips, because they are often composed of dark pixel values, and hence make the corresponding area sunken on the resulting surface. Our main contribution is to resolve this problem by detecting the facial features and making them protrude by adjusting the brightness values of the areas. The experimental results show that our method generates realistic and natural looking bas-relief surfaces that represent more accurate 3D depth, especially in the areas of facial features.

Interactive GPU-based maximum intensity projection of large medical data sets using visibility culling based on the initial occluder and the visible block classification

Maximum intensity projection (MIP) is an important visualization method that has been widely used for the diagnosis of enhanced vessels or bones by rotating or zooming MIP images. With the rapid spread of multidetector-row computed tomography (MDCT) scanners, MDCT scans of a patient generate a large data set. However, previous acceleration methods for MIP rendering of such a data set failed to generate MIP images at interactive rates. In this paper, we propose novel culling methods in both object and image space for interactive MIP rendering of large medical data sets. In object space, for the visibility test of a block, we propose the initial occluder resulting from a preceding image to utilize temporal coherence and increase the block culling ratio a lot. In addition, we propose the hole filling method using the mesh generation and rendering to improve the culling performance during the generation of the initial occluder. In image space, we find out that there is a trade-off between the block culling ratio in object space and the culling efficiency in image space. In this paper, we classify the visible blocks into two types by their visibility. And we propose a balanced culling method by applying a different culling algorithm in image space for each type to utilize the trade-off and improve the rendering speed. Experimental results on twenty CT data sets showed that our method achieved 3.85 times speed up in average without any loss of image quality comparing with conventional bricking method. Using our visibility culling method, we achieved interactive GPU-based MIP rendering of large medical data sets.

A Novel Ranking Model for a Large-Scale Scientific Publication

With a large number of scientific literature, it has been difficult to search for a set of relevant articles and to rank them. In this work, we propose a generalized network analysis approach (called N-star ranking model) for sorting them based on . The ranking of the result is considered in the mutual relationships between another classes: keyword, publication, citation. From the model, we propose two ranks for this problem: the Universal-Publication rank - (UP rank) and Topic-Publication rank (TP rank). We also study two simple ranks based on citation counting (RCC rank) and content matching (RCM rank). We propose the metrics for ranking comparison and analysis on two criteria value and order. We have conducted the experimentations for confirming the predictions and studying the features of the ranks. The results show that the proposed ranks are very impressive for the given problem since they consider the query/topic, the content of publication and the citations in the ranking model.