Kim S J, Kim I J, Bae Y T, Kim Y K, Kim D S
Department of Nuclear Medicine, College of Medicine, Pusan National University, Pusan, Korea.
Acta Radiol. 2005 Apr;46(2):148-54. doi: 10.1080/02841850510020752.
To compare the diagnostic accuracy and incremental diagnostic role of quantitative indices of early and delayed lesion to non-lesion ratios (L/Ns) in the detection of primary breast cancer.
Double-phase (99m)Tc MIBI scintimammography (SMM) (early 10 min, delayed 3 h) was performed after injection of 750 MBq of (99m)Tc MIBI in 446 highly suspected breast cancer patients (malignant: 311, benign: 135). For visual analysis, five scoring methods were used, and, for quantitative analysis, early and delayed L/Ns were calculated. Receiver operating characteristic curve (ROC) analyses were performed to determine the optimal visual grade, and to calculate cut-off values of quantitative indices for differentiation of malignant and benign diseases and to investigate whether the quantitative indices could provide incremental diagnostic values in addition to visual analysis.
Optimal visual grades were above 4 and 5 in the detection of breast cancer. Sensitivity was 84.2% and specificity 79.3%; the area under the curve (AUC) was 0.832 (95% CI, 0.794-0.866) and standard error was 0.019. Early and delayed L/Ns of malignant breast disease were significantly higher than those of benign disease (early: 2.01 +/- 0.99 versus 1.13 +/- 0.26 (P < 0.001); delayed: 1.68 +/- 0.69 versus 1.11 +/- 0.23 (P < 0.001)). The optimal L/Ns for the detection of primary breast cancer were 1.27 for early and 1.12 for delayed imaging. When early L/N 1.27 was used as cut-off value for the detection of primary breast cancer, the sensitivity of SMM was 77.8% and specificity 85.2%. The AUC was 0.856 (95% CI, 0.820-0.888). When delayed L/N 1.12 was used, sensitivity and specificity were 81.4% and 78.5%, respectively. The AUC was 0.834 (95% CI, 0.796-0.867). The ROC comparison of early and delayed L/N showed no statistical difference in the detection of malignant breast disease (P=0.403). When the delayed L/N was added to the early one, early plus delayed quantitative analysis (E+D) showed 86.5% sensitivity and 74.8% specificity. However, the AUCs of E+D (0.854, 95% CI, 0.767-0.842) and early L/N (E) (0.856) showed no statistical difference (P=0.614). When grades 4 and 5 were used as cut-off visual grade, sensitivity and specificity were 84.2% and 79.3%, respectively. When the E was added to visual grade, visual plus early L/N (V+E) showed 89.4% sensitivity and 77% specificity. The AUC of V+E (0.867, 95% CI, 0.832 0.897) was significantly higher than that of visual analysis (V) (0.832, 95% CI, 0.794-0.866, P < 0.001). When the delayed L/N (D) was added to visual grade, visual plus delayed L/N (V+D) showed 89.4% sensitivity and 74.1% specificity. The AUCs of V+D (0.852, 95% CI, 0.816-0.884) and V revealed no statistical differences (P = 0.052).
From this study, the optimal visual grades for diagnosis of breast cancer were grades 4 and 5; the cut-off values of L/Ns were 1.27 for early and 1.12 for delayed imaging. It was also found that early L/Ns provide incremental value in addition to visual analysis. However, delayed L/N revealed no incremental value. Therefore, the delayed image should not be routinely performed for purposes of primary breast cancer detection
比较早期和延迟病变与非病变比值（L/N）的定量指标在原发性乳腺癌检测中的诊断准确性和增量诊断作用。
对446例高度怀疑乳腺癌的患者（恶性：311例，良性：135例）注射750MBq的（99m）Tc MIBI后进行双期（99m）Tc MIBI乳腺闪烁显像（SMM）（早期10分钟，延迟3小时）。视觉分析采用五种评分方法，定量分析计算早期和延迟的L/N。进行受试者操作特征曲线（ROC）分析以确定最佳视觉等级，计算区分恶性和良性疾病的定量指标临界值，并研究定量指标除视觉分析外是否能提供增量诊断价值。
乳腺癌检测的最佳视觉等级在4级及以上。敏感性为84.2%，特异性为79.3%；曲线下面积（AUC）为0.832（95%CI，0.794 - 0.866），标准误为0.019。恶性乳腺疾病的早期和延迟L/N显著高于良性疾病（早期：2.01±0.99对1.13±0.26（P < 0.001）；延迟：1.68±0.69对1.11±0.23（P < 0.001））。原发性乳腺癌检测的最佳L/N，早期为1.27，延迟显像为1.12。当早期L/N 1.27用作原发性乳腺癌检测的临界值时，SMM的敏感性为77.8%，特异性为85.2%。AUC为0.856（95%CI，0.820 - 0.888）。当使用延迟L/N 1.12时，敏感性和特异性分别为81.4%和78.5%。AUC为0.834（95%CI，0.796 - 0.867）。早期和延迟L/N的ROC比较显示在恶性乳腺疾病检测中无统计学差异（P = 0.403）。当将延迟L/N加到早期L/N上时，早期加延迟定量分析（E + D）显示敏感性为86.5%，特异性为74.8%。然而，E + D的AUC（0.854，95%CI，0.767 - 0.842）和早期L/N（E）（0.856）无统计学差异（P = 0.614）。当4级和5级用作临界视觉等级时，敏感性和特异性分别为84.2%和79.3%。当将E加到视觉等级上时，视觉加早期L/N（V + E）显示敏感性为89.4%，特异性为77%。V + E的AUC（0.867，95%CI，0.832 - 0.897）显著高于视觉分析（V）（0.832，95%CI，0.794 - 0.866，P < 0.001）。当将延迟L/N（D）加到视觉等级上时，视觉加延迟L/N（V + D）显示敏感性为89.4%，特异性为74.1%。V + D的AUC（0.852，95%CI，0.816 - 0.884）和V无统计学差异（P = 0.052）。
本研究中，乳腺癌诊断的最佳视觉等级为4级和5级；L/N的临界值，早期为1.27，延迟显像为1.12。还发现早期L/N除视觉分析外提供增量价值。然而，延迟L/N未显示增量价值。因此，对于原发性乳腺癌检测，不应常规进行延迟显像。
