Valverde Francisco L, Guil Nicolás, Muñoz Jose
Department of Computer Science, ETSI Informatica, University of Málaga, Malaga 29071, Spain.
Comput Methods Programs Biomed. 2004 Mar;73(3):233-47. doi: 10.1016/S0169-2607(03)00043-9.
Vessel extraction is a fundamental step in certain medical imaging applications such as angiograms. Different methods are available to segment vessels in medical images, but they are not fully automated (initial vessel points are required) or they are very sensitive to noise in the image. Unfortunately, the presence of noise, the variability of the background, and the low and varying contrast of vessels in many imaging modalities such as mammograms, makes it quite difficult to obtain reliable fully automatic or even semi-automatic vessel detection procedures. In this paper a fully automatic algorithm for the extraction of vessels in noisy medical images is presented and validated for mammograms. The main issue in this research is the negative influence of noise on segmentation algorithms. A two-stage procedure was designed for noise reduction. First, a global approach phase including edge detection and thresholding is applied. Then, the local approach phase performs vessel segmentation using a deformable model with a new energy term that reduces the noise still remaining in the image from the first stage. Experimental results on mammograms show that this method has an excellent performance level in terms of accuracy, sensitivity, and specificity. The computation time also makes it suitable for real-time applications within a clinical environment.
血管提取是某些医学成像应用（如血管造影）中的一个基本步骤。有多种方法可用于在医学图像中分割血管，但它们并非完全自动化（需要初始血管点），或者对图像中的噪声非常敏感。不幸的是，在许多成像模态（如乳腺X光片）中，噪声的存在、背景的变化以及血管的低对比度和对比度变化，使得获得可靠的全自动甚至半自动血管检测程序非常困难。本文提出了一种用于在有噪声的医学图像中提取血管的全自动算法，并在乳腺X光片上进行了验证。本研究的主要问题是噪声对分割算法的负面影响。设计了一个两阶段的降噪程序。首先，应用包括边缘检测和阈值处理的全局方法阶段。然后，局部方法阶段使用具有新能量项的可变形模型进行血管分割，该能量项可减少第一阶段图像中仍残留的噪声。乳腺X光片的实验结果表明，该方法在准确性、敏感性和特异性方面具有优异的性能水平。计算时间也使其适用于临床环境中的实时应用。
