摘要
提出了一种从胸部高分辨率CT图像序列中由粗到细分割肺组织的自动化方法。首先基于结构连续性的分割策略,采用阈值法和数学形态学分割出粗略的肺实质,再用区域生长法去掉气管,最后通过滚球法弥合肺边缘的裂缝及缺口。对6个数据集(共1720层图像)的分割结果成功率均在90%以上,每层分割时间小于6s,相似度分析表明自动与手工分割结果吻合良好,并能较好地保留细节。
An automated lung segmentation method for the serial thoracic high resolution CT images was proposed. Based on the structure contiguity, firstly the method used the thresholding and morphology technique to segment lung parenchyma coarsely. Then it eliminated the trachea and main bronchia using region growing method. Lastly a rolling ball technique was used to remedy the crack and gap in the lung boundary. In experiments, six cases which were comprised of 1720 slices were put into segmentation. The satisfactory results are above 90% and the mean segmentation time of one slice is shorter than 6 seconds. Then a case with normal slices was used to analyze the similarity between the results from computerized segmentation and handwork, and a good superposition is achieved.
出处
《北京生物医学工程》
2008年第1期6-10,共5页
Beijing Biomedical Engineering
基金
安徽省教委重点科研项目(2006KJ097A)资助
