Lesion detectability in digital mammography and digital breast tomosynthesis: A phantom study

Purpose: The complexity of anatomical structure within the breast represents the ultimate limit to signal detection in a mammogram. To increase lesion conspicuity Digital Breast Tomosynthesis (DBT) has been recently proposed and several manufacturers are currently performing clinical trials. In this study we have compared lesion detectability performance of digital mammography and DBT by making use of a phantom in which details of interest are within a heterogeneous background. Methods and Materials: The breast phantom, CIRS (USA) model no. 020 BR3D, consists of various slabs made of heterogeneous tissue-equivalent material that exhibits characteristics of real breast tissue. Between these slabs we have inserted thin layers of homogeneous material containing details of mammographic interest to simulate lesions. A commercial digital mammography unit and a DBT prototype, both manufactured by IMS (Italy), have been used for our study. The 3D reconstruction software is provided by Dexela (UK). 2D and 3D images of the breast phantom have been obtained at various dose levels to compare performance of the two modalities. Results: Comparison between 2D and 3D images recorded at similar dose levels shows superior performance of DBT over digital mammography. Indeed, whilst certain details of interest are not detectable for any dose level with digital mammography, DBT can reveal their signal by reducing complexity of tissue structures. Conclusion: This preliminary investigation demonstrates that in terms of physical image quality the inherent limitations of 2D mammography due to structure noise can be overcome by the introduction of 3D reconstruction via breast tomosynthesis.