摘要
目的 探讨PET图像中勾画感兴趣区 (ROI)的最佳方法及标准摄取值 (SUV)最佳诊断分界点。方法 以PETFDG 12 6例肺部肿瘤图像为例说明 ,由SUVmax的百分比 ( 5 0 %～ 10 0 % )为等高线值勾画ROI ,得各SUV均值 ;由受试者工作特性 (ROC)曲线分析确定ROI最佳画法 ;由准确率最大值、Youden’s指数最大值及敏感性与特异性交点确定分界点。结果 ①最优ROI :70 %等高线所画ROI的SUV具有最大ROC曲线下面积。其 3个典型SUV分界点值分别为 :2 .0 75、2 .82 5、2 .919。② 3个典型分界点的准确率、Youden’s指数和阳性概率比 /阴性概率比均在 70 %等高线出现较大值。结论 提出一套客观画取ROI、确定最佳等高线值、及最佳分界点的方法 。
Objective To investigate the optimum method to draw ROI and the optimum diagnostic cut-off on PET images. Methods PET FDG images of 126 patients with lung tumors (benign 39, malignant 87) were selected. For each tumor six ROIs were drawn with contour line at 50%, 60%, 70%, 80%, 90%, 100% of the maximum SUV, and the average SUVs were calculated. According to the results of ROC analysis, the optimum method of drawing ROI was determined. Three typical diagnostic SUV cut-off points were determined based on the maximum accuracy, the maximum Youden's index and the intersection of sensitivity and specificity. Results ①The area under the ROC curve was maximum at the SUV in the ROI drawn with contour line at 70% of the maximum SUV, and the three typical cut-off SUVs were 2.075, 2.825, 2.919, respectively. ②Among the six ROIs with contour line, the accuracy, Youden's index and the ratio of the positive likelihood ratio (LR+) to the negative likelihood ratio (LR-) at three typical SUVs cut-off had maximum values at the ROI drawn with contour line at 70%. Conclusion The optimum methods of drawing ROI impersonally and determining optimum diagnostic cut-off are proposed, which can be used in quantitative analysis and clinical diagnosis with PET images.
出处
《中国医学影像技术》
CSCD
2004年第6期935-937,共3页
Chinese Journal of Medical Imaging Technology
基金
全军医药卫生科研基金资助项目 (0 1MA1 2 0 )
