Virtual mammogram
Simulating realistic phantom images for mammography
Breast screening relies on X-rays to search for abnormalities in breast tissue in a non-invasive way. The following rule fundamentally applies: the stronger the radiation intensity, the more reliable the diagnosis. However, this also increases the risk of radiation detriment. The settings are optimal when the dose is as low as possible while the image quality remains sufficient. To ensure sufficient imaging quality for the physician to make a diagnosis, devices must be able to represent small structures with low contrast well enough. Within the scope of quality assurance, annual checks are carried out by generating images of a technical phantom that consists of small gold plates of different thicknesses and sizes. The phantom images taken are either assessed by a physician or automatically evaluated by means of procedures of data analysis that simulate human vision.
PTB has developed a simulation software program that simulates the technical phantom and generates realistic phantom images of the gold plates. This virtual mammogram provides images with properties that are exactly known. It therefore allows the analytical procedures used for quality assurance to be validated and optimized. This simulation tool is based on a mathematical model with which a primary image is computed from the transmission of simulated X-rays. Disturbing effects such as scattering or noise are then added. By comparing the generated images with real images, it has been shown that the simulation tool provides realistic mammograms.
This software program has already been used to successfully validate a new assessment procedure that was developed and improved at PTB. In the future, virtual mammograms are to be used in particular to provide a solid basis for the use of machine-learning methods in quality assurance for mammography.
Contact
Tobias Kretz
Departments 8.4, Mathematical Modelling and Data Analysis, and 6.2, Dosimetry for Radiation Therapy and Diagnostic Radiology
Phone: +49 30 3481-7978
tobias.kretz(at)ptb.de
Scientific publication
T. Kretz, M. Anton, T. Schaeffter, C. Elster: Determination of contrastdetail curves in mammography image quality assessment by a parametric model observer, Physica Medica 62, 120–128 (2019)