Dong-Goo Kang

Automatic registration between 3D intra-operative ultrasound and pre-operative CT images of the liver based on robust edge matching

The registration of a three-dimensional (3D) ultrasound (US) image with a computed tomography (CT) or magnetic resonance image is beneficial in various clinical applications such as diagnosis and image-guided intervention of the liver. However, conventional methods usually require a time-consuming and inconvenient manual process for pre-alignment, and the success of this process strongly depends on the proper selection of initial transformation parameters. In this paper, we present an automatic feature-based affine registration procedure of 3D intra-operative US and pre-operative CT images of the liver. In the registration procedure, we first segment vessel lumens and the liver surface from a 3D B-mode US image. We then automatically estimate an initial registration transformation by using the proposed edge matching algorithm. The algorithm finds the most likely correspondences between the vessel centerlines of both images in a non-iterative manner based on a modified Viterbi algorithm. Finally, the registration is iteratively refined on the basis of the global affine transformation by jointly using the vessel and liver surface information. The proposed registration algorithm is validated on synthesized datasets and 20 clinical datasets, through both qualitative and quantitative evaluations. Experimental results show that automatic registration can be successfully achieved between 3D B-mode US and CT images even with a large initial misalignment.

Robust CCD and IR Image Registration Using Gradient-Based Statistical Information

This letter presents a robust similarity measure for registering charged-couple device (CCD) and infrared (IR) images. The measure is based on the entropy obtained from a 3-D joint histogram incorporating edginess and modified generalized gradient vector flow (GGVF). To make a reliable mapping relationship between the edge regions of two images, the concept of edginess was adopted so registration performance would be affected mainly by gradient existences rather than their magnitudes. In addition, by adopting the GGVF, we relaxed a narrow capture range problem in conventional gradient-based measures. Experimental results showed that the proposed measure performs more robustly than existing measures.

A new path planning algorithm for maximizing visibility in computed tomography colonography

In virtual colonoscopy, minimizing the blind areas is important for accurate diagnosis of colonic polyps. Although useful for describing the shape of an object, the centerline is not always the optimal camera path for observing the object. Hence, conventional methods in which the centerline is directly used as a path produce considerable blind areas, especially in areas of high curvature. Our proposed algorithm first approximates the surface of the object by estimating the overall shape and cross-sectional thicknesses. View positions and their corresponding view directions are then jointly determined to enable us to maximally observe the approximated surface. Moreover, by adopting bidirectional navigations, we may reduce the blind area blocked by haustral folds. For comfortable navigation, we carefully smoothen the obtained path and minimize the amount of rotation between consecutive rendered images. For the evaluation, we quantified the overall observable area on the basis of the temporal visibility that reflects the minimum interpretation time of a human observer. The experimental results show that our algorithm improves visibility coverage and also significantly reduces the number of blind areas that have a clinically meaningful size. A sequence of rendered images shows that our algorithm can provide a sequence of centered and comfortable views of colonography.

Real-time head tracking based on color and shape information

In this paper, we propose a robust real-time head tracking algorithm using a pan-tilt-zoom camera. We first assume that the shape of a head is an ellipse and a model color histogram is acquired in advance. Then, in the first frame, the appropriate position and scale of the head is determined based on the user input. In the following frames, the initial position is selected at the same position of the ellipse in the previous frame. Then, the mean shift procedure is applied to make the ellipse position converge to the target center where the color histogram similarity to the model and previous one is maximized. The previous histogram means a color histogram adaptively extracted from the result of the previous frame. Then the position-adjusted ellipse is further refined by using color and shape information. Large background motion often prohibits the initial position from converging to the target position. To alleviate this problem, we estimate a robust initial position by compensating the background motion. Here, to reduce the computational burden of motion estimation, we use vertical and horizontal 1-D projection datasets. Extensive experiments prove that a head is well tracked even when the person moves fast and the scale of the head changes drastically.

Automatic flight path generation in a virtual colonoscopy system

Virtual colonoscopy is a computerized procedure to examine colonic polyps from a CT data set. To automatically fly through a long and complex-shaped colon with a virtual camera, we propose an efficient method to simultaneously generate view-positions and view-directions. After obtaining a 3-D binary colon model, we find an initial path that represents rough camera directions and positions along it. Then, by using this initial path, we generate control planes to find a set of discrete view-positions, and view planes to obtain the corresponding view-directions, respectively. Finally, for continuous and smooth navigation, the obtained view-positions and directions are interpolated using the B-spline method. Here, by imposing a constraint to control planes, penetration and collision can be avoided in the interpolated result. Effectiveness of the proposed algorithm is examined via computer simulations using the several phantoms to simulate the characteristics of human colon, namely, high-curvatures and complex structure. Simulation results show that the algorithm provides the view-positions and view-directions suitable for covering more 3-D surface area in the navigation. Also, prospective results are obtained for human colon data with a high processing speed of less than 1 minute with a 2 GHz standard PC.

Vessel cross-section determination based on nonrigid registration and electric field model

Diameters (or areas) of vessel cross-sections provide useful information for diagnosis and surgery planning. However, the ordinary centerline-perpendicular cross-sections are often inappropriate to use because the centerline may include unwanted local curvatures in irregular or asymmetric regions and high curvatures in sharply bended regions. In this paper, we try to improve the accuracy of vessel cross-section measurement by properly adjusting the centerline. To alleviate local curvatures in the centerline while preserving the global shape faithfully, we register a deformable cylindrical model onto the vessel lumen, and subsequently adopt the axis of the registered model as the adjusted centerline for determining cross-sections. In addition, by introducing the electric field model, we prevent undesirable intersection of cross-sections that is often found in sharply bended regions. Experiments are performed using various synthesized images that simulate abnormal vessels with stenoses or aneurysms. The results show that the registration process successfully eliminates unwanted local curvatures while preserving the global shape of the vessel, and obtained cross-sections do not intersect each other even in the region of high curvature.

A distance-field-based approach in generating cross-sections for 2-D vessel quantification

For accurate determination of thickness-profile in vessel quantification, it is important to find appropriate vessel cross-sections. To obtain vessel cross-sections, a centerline-based approach has been widely used, but it has several inherent problems causing improper cross-sections. First, this approach cannot define cross-sections in a unique way. Second, cross-sections are sensitive to the degree of smoothness of a detected vessel centerline. Third, a small variation in a centerline causes a considerable change in the resultant cross-sections and this phenomenon brings about improper cross-sections in the abnormal vessel of asymmetric structure. Finally, wrong cross-sections may be detected due to the intersection with the other cross-sections in a region of high curvature. In this paper, instead of a centerline, we propose and adopt a complementary geodesic distance field. Then, we detect a sequence of equidistant lines by using the proposed distance field. Finally, we determine cross-sections by refining the obtained equidistant lines. Due to the prospective properties of the proposed distance field, we can alleviate all of the conventional problems and obtain the cross-sections more proper for vessel quantification. Through the intensive simulation using various 2-D synthesized images, we prove that the proposed method provides non-intersecting cross-sections which are insensitive to local variation of geometrical shapes in abnormal vessels.

Automatic path-planning algorithm maximizing observation area for virtual colonoscopy

To navigate a colon lumen, a proper camera path should be generated prior to the navigation. Conventional path-planning algorithms try to find an accurate and robust centerline by assuming that a centerline of colon lumen is the best choice for camera path. For efficient and reliable navigation, however, the centerline may not minimize unobservable area from the camera path. In this paper, we first define a new coverage measure reflecting the temporal visibility. And based on this measure, a fast and efficient path-planning algorithm is proposed to increase the visibility coverage. The proposed algorithm first simplifies the object surface using the centerline determined. Then, camera view positions and directions are estimated to maximize the observable surface. Simulation results prove that the proposed algorithm provides a better coverage rate than the conventional one without a significant increase of additional computation.