| 何茜,贺文,王克扬,马大庆,王涛,李晨,田扬.数字胸片多频域后处理对计算机辅助检测系统肺结节检测的影响[J].中国医学影像技术,2007,23(12):1888~1891 |
| 数字胸片多频域后处理对计算机辅助检测系统肺结节检测的影响 |
| Effect of multiscale processing in digital chest radiography oncomputed-aided lung nodule detection |
| 投稿时间：2007-07-12 修订日期：2007-08-23 |
| DOI： |
| 中文关键词: 数字影像 多频域后处理 肺结节 计算机辅助检测 ROC曲线 |
| 英文关键词:Digital radiography Multiscale processing Lung nodule Computed-aided diagnosis Receiver operating characteristic curve |
| 基金项目: |
|
| 摘要点击次数: 2270 |
| 全文下载次数: 1364 |
| 中文摘要: |
| 目的 评价数字胸片多频域后处理对计算机辅助检测(CAD)系统肺结节检出的影响。方法 对经CT证实的54例肺结节病例及54例正常者的数字胸片选择三种不同的structure preference参数进行多频域后处理,在1次曝光的条件下各得到标准图像、高通过图像及低通过图像三组图像,然后,采用计算机辅助检测系统(IQQATM-Chest V1.2)进行阅片,首先由2名观察者共同根据CT结果客观分析并记录CAD检出的结节数及假阳性数。其次,由另外4名观察者(高年资及低年资放射医师各2名)应用CAD系统所提供的肺结节的智能质量和数量分析功能,对CAD 输出图像独立进行分析,并记录检出结节数及假阳性数,采用受试者操作特征(ROC)曲线分析观察结果。结果 三组图像中,低通过组图像的平均ROC曲线下面积最大,人-机交互前、高年资及低年资人-机交互后的平均ROC曲线下面积分别为0.77、0.81、0.80。高通过组图像的平均ROC曲线下面积为最小,人-机交互前、高年资及低年资人-机交互后的平均ROC曲线下面积分别为0.57、0.67、0.71。三组图像的平均ROC曲线下面积存在差异(P<0.01)。结论 数字胸片多频域后处理影响计算机辅助检测系统肺结节的检出。 |
| 英文摘要: |
| Objective To evaluate the effect of multiscale processing in digital chest radiography on lung nodule detection with a computer-aided diagnosis (CAD) system. Methods The study involved fifty-four small nodule patient cases and fifty-four normal cases on the basis of confirmation on chest CT. Each case underwent an examination by means of one time X-ray exposing with a digital radiography system, and the acquired image was processed by three types of multiscale processing with different structure preference values. Two experienced radiologists established the nodule gold standard by consensus reading according to CT results, and analyzed and recorded the detection of lung nodules and false positive detections of these CAD output images. Then, two less experienced radiologists and other two experienced radiologists independently analyzed and recorded the detection of lung nodules and false positive detections of these images using the intelligent qualitative and quantitative analysis provided by the CAD system. All observations were evaluated by the receiver operating characteristic(ROC)analysis. Results Among the three kinds of images, the mean areas under the ROC curve (Az) values of the low-pass images were highest, and Az values were 0.77, 0.81, 0.80 for the low-pass images before human-computer interact, experienced radiologists and less experienced radiologists, respectively. The Az values of the high-pass images were lowest, and Az values were 0.57, 0.67, 0.71 for the low-pass images before interact, experienced radiologists and less experienced radiologists, respectively. There were statistical significances of the mean area under the receiver operating characteristic curve values among these three kinds of processed images (P<0.01). Conclusion Multiscale processing in digital chest radiography affects the lung nodule detection by the CAD. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
