Multiscale Boundary Identification for Medical Images
Boundary identification in medical images plays a crucial role in helping physicians in patient diagnosis. Manual identification of object boundaries is a time-consuming task and is subject to operator variability. Fully automatic procedures are still far
- PDF / 549,476 Bytes
- 9 Pages / 430 x 660 pts Page_size
- 32 Downloads / 200 Views
Abstract. Boundary identification in medical images plays a crucial role in helping physicians in patient diagnosis. Manual identification of object boundaries is a time-consuming task and is subject to operator variability. Fully automatic procedures are still far from satisfactory in most real situations. In this paper, we propose a boundary identification method based on multiscale technique. Experimental results have shown that the proposed method provides superior performance in medical image segmentation. Keywords: Boundary identification, multiscale edge detection.
1
Introduction
Various medical imaging modalities such as the radiograph, computed tomography (CT), and magnetic resonance (MR) imaging are widely used in routine clinical practice. Boundary identification (BI) of deformed tissue plays a crucial role in accurate patient diagnosis. For example, an MR brain image can be segmented into different tissue classes, such as gray matter, white matter, and cerebrospinal fluid. Unfortunately, manual BI methods are very time consuming and are often subjective in nature. Recently an edge-based MR image segmentation method, mtrack [1], has been proposed. This algorithm provides a reasonably good performance. However, the edge-based segmentation algorithm has some limitations, such as sensitivity to noise and presence of gaps in detected boundaries. Together, these effects degrade the quality of the detected boundaries. The discrete wavelet transform (DWT) has recently been shown to be a powerful tool in multiscale edge detection. Marr and Hildreth [2] introduced the concept of multiscale edge detection for detecting the boundaries of objects in an image. Mallat and Zhong [3] showed that multiscale edge detection can be implemented by smoothing a signal with a convolution kernel at various scales, and detecting sharp variation points as edges. Tang et al. [4] studied the characterization of edges with wavelet transform and Lipschitz exponents. In this paper, we propose a multiscale boundary identification algorithm, based on DWT, for better edge-based segmentation of MR images. The remainder of this paper is organized as follows. Section 2 briefly reviews the related background work. Section 3 presents the proposed method developed based on A. Ghosh, R.K. De, and S.K. Pal (Eds.): PReMI 2007, LNCS 4815, pp. 177–185, 2007. c Springer-Verlag Berlin Heidelberg 2007
178
Z. Wang, M. Mandal, and Z. Koles
the existing theory for multiscale edge detection and boundary identification. Section 4 presents the performance evaluation of the proposed method. The conclusions are presented in Section 5.
2
Review of Related Work
In this section, we present a brief review of the related background work. 2.1
mtrack Algorithm
The mtrack method [1] has been developed at the University of Alberta for medical image segmentation, especially for MR images of the head. As shown in Fig. 1, it includes three major modules: edge detection, edge feature extraction, and edge tracing. The three modules are explained below.
Fig. 1. Block diagr
Data Loading...
