应用ROC曲线分析钙化点分布及形态特征对乳腺癌诊断的价值

目的：探讨应用ROC曲线分析钙化点分布及形态特征对乳腺癌诊断的价值。方法：收集由超声影像检查发现钙化点并最终经病理诊断为乳腺癌的198例患者的影像资料，通过ROC曲线分析钙化点分布情况及其形态特征，包括单位面积个数、不同形态钙化的概率、区域平均直径、最长直径的差异等。采用ROC曲线分析数据，了解其对乳腺癌的诊断价值，并确认其最佳临界值。结果：恶性病变组中钙化点分布最佳单位面积个数是34个／m2，区域平均直径22．98mm，长径差异1．41ram，沙砾样钙化率占78．9％，线样分支状钙化率51．2％；ROC曲线下面积各为：0．745、0．756、0．678、0．860、0．900。其中线样分支状钙化率诊断准确率最高。结论：各种钙化分布的特征均有诊断效能，但单一的因素存在局限性，综合检测多项特征能有效的提高准确率。

The Application of ROC Curve on the Analysis of Calcification Point Distribution and Morphological Character for the Diagnosis of Breast Cancer

Objective： To investigate the application of ROC curve on the analysis of calcification point distribution and morphological character for the diagnosis of breast cancer. Methods： 198 patients＇ image data were collected by molybdenum target imaging findings of calcification point and the final pathological diagnosis that showed the breast cancer. Then the ROC curve was used to analyze the calcification point distribution and morphological character, which including the per unit area number, probability of different forms of calcification, the regional average diameter, the longest diameter＇s difference, and the optimal critical value was confirmed. Results： The malignant group of calcification in the best distribution unit area number was 34 per square meter, Regional average diameter was 22.98mm, Long diameter difference was 1.41mm, Gravel like calcification rate of 78.9%, Line like branching calcification rate was 51.2%; The area under the ROC curve of each： 0.745, 0.756, 0.678, 0.860, 0.900.The line like branching calcification＇s diagnostic accuracy rate was the highest. Conclusion： The various calcification distribution features have diagnostic efficacy, except of the limitations with single factor. Detection of multiple features can effectively improve the accuracy.