Development of Computer-Aided Diagnosis System in Screening Digital Chest Radiographs

为了提高肺部结节的检出率，我开发了一个计算机辅助诊断系统（CAD），用于数字化胸部X线摄影的筛查。CAD指的是放射科医生将计算机自动分析放射图像的结果作为“第二意见”加以考虑的诊断。“在这个项目中，我改进了以前开发的一种计算机化的肺结节检测方法，并对时间减影技术进行了初步的研究，该技术可以通过从当前图像减去先前图像来增强间隔变化。1.推导了一种新的检测算法，该算法通过使用匹配滤波器来增加不同大小肺结节的对比度。我将此方法应用于200张有肺结节的胸片，以研究检测的准确性。在1.7个假阳性/图像下，检测性能为79%的真阳性。2.我推导了一种时间减法技术，用于匹配先前和当前图像。我应用这种技术的876对以前和当前的图像获得的移动的计算机X射线摄影（CR）系统。结果表明，该系统能有效地检测出肺部结节，为临床诊断提供了一种新的方法。

I have developed a system of computer-aided diagnosis (CAD) in screening digital chest radiography for improvement of detection accuracy of lung nodules. CAD refers to a diagnosis by a radiologist taking into consideration the results of an automated computer analysis of radiographic images as a "second opinion." In this project, I improved a computerized detection method of lung nodules, which was developed previously.In addition, I performed a preliminary research of temporal subtraction technique which can enhance the interval changes by subtracting a previous image from a current one.1.I derived a new detection algorithm which increases the contrast of lung nodules with various sizes by using matched filters. I applied this method for 200 chest radiographs with lung nodules in order to investigate the detection accuracy. The detection performance was 79% of true positive at 1.7 false positives/image.2.I derived a temporal subtraction technique with traslation, rotation and warping for matching previous and current images. I applied this technique for 876 pairs of previous and current images obtained by a mobile computed radiography (CR) system. 76.7% of subtraction images without misregistration artifacts were successfully provided.3.I provided an image database including approximately 10,000 pairs of previous and current images during the term of this project.4.My results indicates that this CAD system has a potential usefulness for radiologists to detect lung nodules in the screening chest radiography.