正电子发射乳腺断层成像术（PEM）：活性浓度、物体大小和物体对比度对图像伪影检测的影响。

Positron emission mammography (PEM): effect of activity concentration, object size, and object contrast on phantom lesion detection.

机构信息

Department of Radiology, University of Washington, Seattle, WA 98195, USA.

出版信息

Med Phys. 2012 Oct;39(10):6499-508. doi: 10.1118/1.4754651.

DOI:10.1118/1.4754651

PMID:23039684

Abstract

PURPOSE

To characterize the relationship between lesion detection sensitivity and injected activity as a function of lesion size and contrast on the PEM (positron emission mammography) Flex Solo II scanner using phantom experiments.

METHODS

Phantom lesions (spheres 4, 8, 12, 16, and 20 mm diameter) were randomly located in uniform background. Sphere activity concentrations were 3 to 21 times the background activity concentration (BGc). BGc was a surrogate for injected activity; BGc ranged from 0.44-4.1 kBq∕mL, corresponding to 46-400 MBq injections. Seven radiologists read 108 images containing zero, one, or two spheres. Readers used a 5-point confidence scale to score the presence of spheres.

RESULTS

Sensitivity was 100% for lesions ≥12 mm under all conditions except for one 12 mm sphere with the lowest contrast and lowest BGc (60% sensitivity). Sensitivity was 100% for 8 mm spheres when either contrast or BGc was high, and 100% for 4 mm spheres only when both contrast and BGc were highest. Sphere contrast recovery coefficients (CRC) were 49%, 34%, 26%, 14%, and 2.8% for the largest to smallest spheres. Cumulative specificity was 98%.

CONCLUSIONS

Phantom lesion detection sensitivity depends more on sphere size and contrast than on BGc. Detection sensitivity remained ≥90% for injected activities as low as 100 MBq, for lesions ≥8 mm. Low CRC in 4 mm objects results in moderate detection sensitivity even for 400 MBq injected activity, making it impractical to optimize injected activity for such lesions. Low CRC indicates that when lesions <8 mm are observed on PEM images they are highly tracer avid with greater potential of clinical significance. High specificity (98%) suggests that image statistical noise does not lead to false positive findings. These results apply to the 85 mm thick object used to obtain them; lesion detectability should be better (worse) for thinner (thicker) objects based on the reduced (increased) influence of photon attenuation.

摘要

目的

使用体模实验，研究在 PEM（正电子发射乳腺成像）Flex Solo II 扫描仪上，作为病变大小和对比度函数的病变检测灵敏度与注射活动之间的关系。

方法

体模病变（直径 4、8、12、16 和 20 毫米的球体）随机分布在均匀背景中。球体的活性浓度是背景活性浓度（BGc）的 3 到 21 倍。BGc 是注射活动的替代物；BGc 范围为 0.44-4.1 kBq∕mL，对应于 46-400 MBq 的注射。七位放射科医生阅读了包含零、一个或两个球体的 108 张图像。读者使用 5 分置信度评分来判断球体的存在。

结果

除了一个具有最低对比度和最低 BGc（60%灵敏度）的 12 毫米球体之外，所有条件下直径≥12 毫米的病变的灵敏度均为 100%。当对比度或 BGc 较高时，8 毫米球体的灵敏度为 100%，只有当对比度和 BGc 均最高时，4 毫米球体的灵敏度为 100%。最大到最小球体的球体对比度恢复系数（CRC）分别为 49%、34%、26%、14%和 2.8%。累积特异性为 98%。

结论

体模病变检测灵敏度更多地取决于球体大小和对比度，而不是 BGc。对于低至 100 MBq 的注射活动，灵敏度仍保持≥90%，对于直径≥8 毫米的病变。4 毫米物体的低 CRC 导致即使对于 400 MBq 的注射活动，检测灵敏度也适中，因此对于此类病变优化注射活动是不切实际的。低 CRC 表明，当在 PEM 图像上观察到<8 毫米的病变时，它们具有较高的示踪剂亲和力，具有更大的临床意义的可能性。高特异性（98%）表明图像统计噪声不会导致假阳性发现。这些结果适用于获得它们的 85 毫米厚的物体；基于光子衰减的减少（增加），对于更薄（更厚）的物体，病变的可检测性应该更好（更差）。