Structural Edge Detection for Cardiovascular Modeling
Computational simulations provide detailed hemodynamics and physiological data that can assist in clinical decision-making. However, accurate cardiovascular simulations require complete 3D models constructed from image data. Though edge localization is a
- PDF / 1,232,815 Bytes
- 8 Pages / 439.363 x 666.131 pts Page_size
- 119 Downloads / 209 Views
4
Electrical and Computer Engineering 2 Cognitive Science 3 Computer Science and Engineering Mechanical and Aerospace Engineering University of California, San Diego
Abstract. Computational simulations provide detailed hemodynamics and physiological data that can assist in clinical decision-making. However, accurate cardiovascular simulations require complete 3D models constructed from image data. Though edge localization is a key aspect in pinpointing vessel walls in many segmentation tools, the edge detection algorithms widely utilized by the medical imaging community have remained static. In this paper, we propose a novel approach to medical image edge detection by adopting the powerful structured forest detector and extending its application to the medical imaging domain. First, we specify an effective set of medical imaging driven features. Second, we directly incorporate an adaptive prior to create a robust three-dimensional edge classifier. Last, we boost our accuracy through an intelligent sampling scheme that only samples areas of importance to edge fidelity. Through experimentation, we demonstrate that the proposed method outperforms widely used edge detectors and probabilistic boosting tree edge classifiers and is robust to error in a prori information.
1
Introduction
Building on advances in medical imaging technology, cardiovascular blood flow simulation has emerged as a non-invasive and low risk method to provide detailed hemodynamic data and predictive capabilities that imaging alone cannot [4]. A necessary precursor to these simulations is accurate construction of patient-specific 3D models via segmentation of medical image data. Currently available tools to aid model construction include ITK-SNAP (http://www.itksnap.org) for general medical image segmentation and SimVascular (http://www.simvascular.org) and vmtk (http://www.vmtk.org) which use specialized segmentation techniques for blood flow simulation. These packages implement automated segmentation tools, most of which rely heavily on region growers, active-contours and snakes. These methods can be effective for cardiovascular segmentation, however they often require finely-tuned algorithm parameters, hence manual segmentation is commonly used in practice. Model creation remains one of the main bottle-necks in widespread simulation c Springer International Publishing Switzerland 2015 N. Navab et al. (Eds.): MICCAI 2015, Part III, LNCS 9351, pp. 735–742, 2015. DOI: 10.1007/978-3-319-24574-4_88
736
J. Merkow et al.
of patient specific cardiovascular systems. Increased efficiency of the model construction process will enable simulations in larger cohorts of patients, increasing the clinical impact of simulation tools. Machine learning techniques produce robust results without manual interaction, which make them a powerful tool for model construction. There has been extensive research into learned edge detection in natural images but edge detection has not received the same attention in medical imaging. In this paper, a machine learning based edge de
Data Loading...
