Performance assessment of breast tomosynthesis systems: Concepts for two types of phantoms

Abstract

In order to justify the clinical use of a new modality such as digital breast tomosynthesis (DBT), a performance test in terms of a clinical parameter is indicated. For screening applications, lesion detectability is crucial. Whereas a real phantom with typically a homogeneous background that includes a series of details may provide a good indication of such detectability, a more specific approach is needed if the potential advantages of DBT have to be assessed. More specifically, the potential of the reduction of soft tissue superposition has to be covered in the performance test. We propose 2 concepts for a performance evaluation phantom: a hardware phantom with lesion-like inserts and where a background can be simulated on top of the images, and a software phantom with simulated lesions of different size and density (that can originate from theoretical models or from attenuation templates of real lesions). We discuss the need and the use of these phantoms for performance testing. © 2010 Springer-Verlag.