SU‐FF‐I‐109: Quantitative Breast Tomosynthesis: Development of An Estimation Performance Metric and Optimization Framework

Purpose: To develop performance metrics for estimation tasks and employ the framework to optimize breast tomosynthesis systems for quantitative imaging purposes. Method and Materials: A maximum likelihood (ML) estimator was derived in terms of the noise power spectrum (NPS), which yielded a figure of merit for quantitative imaging performance denoted the ML mean error (sML). The sML was computed for the estimation of lesion diameter, volume, and attenuation. Volumetric tomosynthesis reconstructions of a breast phantom, which incorporated electronic, quantum, and anatomical noise (1/fb) with embedded spherical lesions were simulated at tomosynthesis acquisition angles (tomo‐angle) varying from 4o to 200o and at constant total acquisition dose (1.5 mGy). The estimation task results were further compared with a more conventional lesion detection task. Results: Results reveals tradeoffs between electronic, quantum, and anatomical noise. For a 1.25 mm radius spherical lesion, the mean error (sML) as a function of tomo‐angle varied from 0.06 to 0.5 mm for the 2D disk radius estimation task and from 0.2 to 0.9 mm for the 3D sphere radius estimation task. Furthermore, for a given sML, the 3D size estimation task required a larger tomo‐angle than for the 2D size estimation task, highlighting the need for larger acquisition angles in 3D imaging tasks (i.e., volume estimation). Overall, size estimation performance was maximized at a tomo‐angle of ∼90o. Results also suggested that size estimation tasks generally required a larger tomo‐angle than conventional detection task. Conclusions: Analysis of quantitative imaging performance using Fourier metrics highlights the difference between 2D and 3D estimation tasks in breast tomosynthesis and provides a meaningful framework for optimizing the performance of volumetric imaging systems for quantitative imaging tasks. © 2009, American Association of Physicists in Medicine. All rights reserved.

Duke Scholars

Author Ehsan Samei Radiology