Anatomical and quantum noise inhibits detection of malignancies in clinical images such as in digital mammography (DM), digital breast tomosynthesis (DBT) and breast CT (bCT). In this work, we examine the relative influence and interactions of these two types of noise on the task of low contrast mass detectability in DBT. We show how the changing levels of quantum noise contributes to the estimated power-law slope $\beta $ by changing DBT acquisition parameters as well as with spatial filtering like an adaptive Weiner filtering. Finally, we examine via human observer LROC studies whether power spectral parameters obtained from DBT images correlate with mass detectability in those images. Our results show that lower values of power-law slope $\beta $ can result from heightened quantum noise or image artifacts and do not necessarily imply reduced anatomical noise or improved signal detectability for the given imaging system. These results strengthen the argument that when power-law magnitude K is varying, $\beta $ is less relevant to lesion detectability. Our preliminary results also point to K values having strong correlation to human observer performance, at least for the task shown in this paper. As a byproduct of these main results, we also show that while changes in acquisition geometry can improve mass detectability, the use of efficient filters like an adaptive Weiner filtering can significantly improve the detection of low contrast masses in DBT.

中文翻译：

---

解剖学和量子噪声在信号检测和断层扫描数字乳房图像感知中的相对贡献。

解剖学和量子噪声抑制了临床图像（例如，乳腺X线摄影（DM），乳腺X线断层合成（DBT）和乳腺CT（bCT））中恶性肿瘤的检测。在这项工作中，我们研究了这两种类型的噪声对DBT中低对比度质量可检测性任务的相对影响和相互作用。我们展示了量子噪声水平的变化如何对估计的幂律斜率做出贡献 $ \ beta $ 通过更改DBT采集参数以及像自适应Weiner滤波这样的空间滤波。最后，我们通过人类观察者LROC研究检查了从DBT图像获得的功率谱参数是否与这些图像中的质量检测能力相关。我们的结果表明，幂律斜率的值较低 $ \ beta $ 量子噪声或图像伪影的增加可能会导致这种现象，并且对于给定的成像系统，不一定意味着解剖噪声降低或信号检测能力提高。这些结果进一步证明了当幂律幅度K发生变化时， $ \ beta $ 与病变的可检测性无关。我们的初步结果还指出，至少对于本文所示的任务，K值与人类观察者的表现有很强的相关性。作为这些主要结果的副产品，我们还表明，虽然采集几何形状的变化可以改善质量检测能力，但使用像自适应Weiner滤波这样的有效滤波器可以显着改善DBT中低对比度质量的检测。