Sanada S, Doi K, MacMahon H
Kurt Rossmann Laboratories for Radiologic Image Research, Department of Radiology, University of Chicago, Illinois 60637.
Med Phys. 1991 Sep-Oct;18(5):964-71. doi: 10.1118/1.596611.
In order to facilitate computerized quantitative analysis of digital chest radiographs, an automated method for accurate delineation of posterior ribs in frontal chest images is being developed. This method is based on an analysis of vertical profiles in the lung regions and a statistical analysis of edge gradients and their orientations in small selected regions-of-interest (ROIs). A shift-variant function is fitted to vertical profiles to obtain initial estimates of locations of rib edges. Rib edges are then determined more accurately by analyzing cumulative edge gradients and their orientations in small ROIs that are located adjacent to the initially estimated edges. The present computerized method can achieve a good agreement between the detected and the actual rib structures for posterior ribs in 74% of 50 cases examined. This suggests that automated detection of posterior ribs by a computerized method is feasible, and may be useful for computer-aided diagnostic schemes in the chest.
为便于对数字化胸部X光片进行计算机定量分析,正在开发一种用于在胸部正位图像中准确描绘后肋的自动化方法。该方法基于对肺部区域垂直轮廓的分析以及对小的选定感兴趣区域(ROI)中边缘梯度及其方向的统计分析。将一个移位可变函数拟合到垂直轮廓上,以获得肋骨边缘位置的初始估计值。然后,通过分析位于初始估计边缘附近的小ROI中的累积边缘梯度及其方向,更准确地确定肋骨边缘。在所检查的50例病例中,有74%的病例采用当前的计算机方法能够使检测到的后肋与实际肋骨结构达到良好的一致性。这表明通过计算机方法自动检测后肋是可行的,并且可能对胸部的计算机辅助诊断方案有用。