Dose Reduction and Performance Enhancement During DBT Screening

描述（由申请人提供）：数字乳腺断层合成（DBT）能够从不同角度获得的一系列二维投影中对多层乳腺组织进行三维重建。它的主要优点是，它生成了一个体积数据集，允许组织重叠模糊的分辨率，经常出现在投影乳房x光检查。目前正在研究DBT作为主要筛查方式的使用，但与辐射暴露和检测性能有关的问题仍然存在。在之前的回顾性观察中，我们发现DBT联合全场数字乳房x线摄影（FFDM）在保持灵敏度的同时，与单独的FFDM相比，召回率降低了30%，但联合检查使辐射暴露增加了一倍。当DBT单独与FFDM进行比较时，减少的仅为10%。召回率降低30%足以证明将DBT纳入筛查方案是合理的，但如果它需要比FFDM增加一倍的辐射暴露，就不合理了。本提案的目的是确定是否可以在不单独获取FFDM和DBT研究的情况下，通过使用合成FFDM（由适当重建的DBT数据的高分辨率（70-5m）投影组成）代替直接获取的FFDM来实现30%的减少。为了验证这一点，我们将开发从DBT数据中导出合成FFDM的算法，然后进行观察者性能研究，测量1)FFDM单独的性能，然后FFDM结合传统DBT，以及2)传统DBT单独的性能，然后DBT结合合成FFDM。模式1先前已被证明可以提供上述讨论的性能增益，但需要模式2的两倍辐射剂量。这两个研究臂使我们能够单独测量FFDM、FFDM + DBT、DBT单独和DBT +合成FFDM的性能。表现研究将有8名放射科医生阅读100例两种模式下的病例，并报告BIRADS评分和roc型评分。每个病例将包括在Hologic DBT系统上通过单一程序获得的FFDM和DBT。这项工作将是重要的，因为如果我们的目标是用DBT数据衍生的投影取代FFDM采集，那么就有可能在不增加辐射暴露的情况下实现召回率降低30%。

DESCRIPTION (provided by applicant): Digital Breast Tomosynthesis (DBT) enables 3-D reconstruction of multiple levels of breast tissue from a series of 2-D projections acquired at different angles. Its main advantage is that it generates a volumetric dataset that allows resolution of tissue superposition ambiguities that are often present in projection mammograms. DBT is currently being investigated for use as a primary screening modality but issues related to radiation exposure and detection performance remain. In a previous retrospective observer performance we found that DBT in combination with Full Field Digital Mammography (FFDM) reduced recall rates by 30% while maintaining sensitivity, compared to FFDM alone, but the combined examination doubled radiation exposure. When DBT alone was compared to FFDM the reduction was only 10%. A 30% reduction in recall rates would be sufficient to justify incorporating DBT into screening protocols, but not if it requires doubling the radiation exposure over that of FFDM. The intent of this proposal is to determine whether the 30% reduction can be achieved without acquiring separate FFDM and DBT studies by using a synthetic FFDM, comprised of a high-resolution (70-5m) projection of appropriately reconstructed DBT data, as a substitute for the directly acquired FFDM. To test this we will develop algorithms for deriving the synthetic FFDM from the DBT data and then perform an observer performance study that measures performance of 1) FFDM alone followed by FFDM combined with traditional DBT, and 2) traditional DBT alone followed by DBT combined with the synthetic FFDM. Mode 1 has previously been shown to provide the performance gains discussed above, but requires twice the radiation dose of mode 2. The two study arms enable us to measure performance of FFDM alone, FFDM + DBT, DBT alone, and DBT + synthetic FFDM. The performance study will have 8 radiologists read 100 cases in both modes and report BIRADS scores and ROC-type scores. Each case will include an FFDM and a DBT acquired in a single procedure on our Hologic DBT system. This work will be significant because, if our goal of replacing the FFDM acquisition with a projection derived from the DBT data is successful, then it will be possible to achieve the 30% reduction in recall rates without the increased radiation exposure. PUBLIC HEALTH RELEVANCE: Breast tomosynthesis is a recently introduced imaging modality that can significantly reduce recall rates, while maintaining screening sensitivity, if used in conjunction with mammography in a screening environment. However, this combined exam doubles the radiation exposure over that required for mammography alone. This project tests various strategies for incorporating DBT into breast cancer screening environments that may be able to retain the performance improvements without the increase in radiation exposure. This could greatly reduce the number of women unnecessarily recalled because of a false positive screening result.