Centerline-based analyses of complex-shaped tubular structures in 3-D medical images

Abstract

As medical imaging and computer technologies advances rapidly, computerized analysis of medical images is becoming increasingly important. In particular, analysis of tubular objects in a human body is of concern to various medical applications. In many algorithms for dealing with tubular objects, a centerline has been widely en-gaged as an important tool due to its efficient expressivity of tubular structures in qualitative and quantitative analyses. Inappropriate use of a centerline, however, often leads to performance degradation. This thesis presents two novel centerline-based approaches aiming at two medical applications: Virtual colonoscopy and vascular abnormality quantification. First, in virtual colonoscopy, or a computer-aided screen-ing modality for early detection of colonic polyps, the minimization of blind areas is a critical issue for accurate and effective diagnosis. Although the centerline is useful for describing the shape of an object, it does not always provide the optimal fly-through path in terms of minimum blind area. In practice, considerable blind areas can be produced by adopting a centerline as a path, especially in areas of high curvature. This thesis presents a novel centerline-based approach to the minimization of blind areas by incorporating surface information into a path-planning process. Contrary to conven-tional schemes, view positions and their corresponding view directions are jointly determined to increase the visibility of surface area. In addition, a new iterative path-smoothening scheme is proposed to provide comfortable views without degradation of visibility performance. For the evaluation of the proposed algorithm, we first attempt to quantify the overall observable area on the basis of the temporal visibility measure that quantifies the minimum interpretation time of a human observer. The experimen-tal results show that proposed algorithm improves visibility coverage and also signifi-cantly reduces the number of blind...