The authors compared the performance of five protocols intended to reduce dose to the breast during computed tomography (CT) coronary angiography scans using a model observer unknown-location signal-detectability metric.
The authors simulated CT images of an anthropomorphic female thorax phantom for a 120 kV reference protocol and five “dose reduction” protocols intended to reduce dose to the breast: 120 kV partial angle (posteriorly centered), 120 kV tube-current modulated (TCM), 120 kV with shielded breasts, 80 kV, and 80 kV partial angle (posteriorly centered). Two image quality tasks were investigated: the detection and localization of 4-mm, 3.25 mg/ml and 1-mm, 6.0 mg/ml iodine contrast signals randomly located in the heart region. For each protocol, the authors plotted the signal detectability, as quantified by the area under the exponentially transformed free response characteristic curve estimator ( ), as well as noise and contrast-to-noise ratio (CNR) versus breast and lung dose. In addition, the authors quantified each protocol's dose performance as the percent difference in dose relative to the reference protocol achieved while maintaining equivalent .
For the 4-mm signal-size task, the 80 kV full scan and 80 kV partial angle protocols decreased dose to the breast (80.5% and 85.3%, respectively) and lung (80.5% and 76.7%, respectively) with = 0.96, but also resulted in an approximate three-fold increase in image noise. The 120 kV partial protocol reduced dose to the breast (17.6%) at the expense of increased lung dose (25.3%). The TCM algorithm decreased dose to the breast (6.0%) and lung (10.4%). Breast shielding increased breast dose (67.8%) and lung dose (103.4%). The 80 kV and 80 kV partial protocols demonstrated greater dose reductions for the 4-mm task than for the 1-mm task, and the shielded protocol showed a larger increase in dose for the 4-mm task than for the 1-mm task. In general, the CNR curves indicate a similar relative ranking of protocol performance as the corresponding curves, however, the CNR metric overestimated the performance of the shielded protocol for both tasks, leading to corresponding underestimates in the relative dose increases compared to those obtained when using the metric.
The 80 kV and 80 kV partial angle protocols demonstrated the greatest reduction to breast and lung dose, however, the subsequent increase in image noise may be deemed clinically unacceptable. Tube output for these protocols can be adjusted to achieve a more desirable noise level with lesser breast dose savings. Breast shielding increased breast and lung dose when maintaining equivalent . The results demonstrated that comparisons of dose performance depend on both the image quality metric and the specific task, and that CNR may not be a reliable metric of signal detectability.
This study was supported in part by an appointment to the Research Participation Program at the FDA Center for Devices and Radiological Health administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the United States Department of Energy and the Food and Drug Administration, Office of Women's Health. Computer simulations were performed on the Marquette University Pere High Performance Computing Cluster, which was funded in part by National Science Foundation Award Nos. OCI-0923037 and CBET-0521602. The authors would like to thank Lars E. Olson, Ph.D. and David Herzfeld for their support with the high performance cluster.
II.B. Unknown-location signal-detectability metric () estimation
II.C. Simulation setup
II.D. Simulation geometry
II.E. Investigated protocols
II.F. Dose estimation
II.G. Image generation
II.H. Assessment of dose and image quality
II.H.1. Signal detectability,
II.H.2. Dose reduction at equivalent
II.H.3. Pixel noise,contrast, and contrast-to-noise ratio estimation
- Medical imaging
- Medical image noise
- Medical image quality
- Computed tomography
Data & Media loading...
Article metrics loading...
Full text loading...