Jurgen Jacobs MSc, Abstract Co-Author: Nothing to Disclose

Nicholas William Marshall BSC, PhD, Abstract Co-Author: Nothing to Disclose

Guy Jacques Marchal MD, PhD, Abstract Co-Author: Nothing to Disclose

Hilde Bosmans PhD, Presenter: Nothing to Disclose

To report our first experience with a newly developed software framework for automated and remote quality control (QC) of digital breast tomosynthesis (DBT) systems using image summaries.

We developed a vendor independent analysis method to perform QC of DBT systems remotely. Two phantoms have been used: a homogeneous phantom and an experimental phantom containing elements to check for missing breast tissue, reconstruction artifacts and uniformity. The current software solution could easily be extended to support other phantoms. A number of parameters (mean pixel value, signal-to-noise ratio, standard deviation, variance, … ) are calculated from a series of small overlapping regions of interest (ROI) in each projective image and a set of tomographic planes. Starting from these data, thumbnail images, color deviation maps and difference maps against a reference slice are generated. Additional calculated values are noise power spectra (NPS) and contrast-to-noise ratio (CNR). The artifact spread function (ASF) can be calculated in detected artifacts. Selected DICOM header values are recorded for follow-up in time. To save bandwidth, only the image analysis data packed in a dedicated file format need to be sent to a reference center. This system has been tested using data from 3 prototype (GE Healthcare Senographe DS modified for tomosynthesis investigation; IMS Giotto prototype,Bologna,Italy; Siemens MAMMOMAT NovationTOMO prototype) and 1 commercially available DBT system (Hologic Selenia Dimensions). The whole analysis chain is optimized to use multi processor computer architectures.

The implementation of our software was successful for each tested system. Per 100 MB input data the analysis typically takes 54 seconds and generates 9.33MB analysis results on a computer with a quad-core processor. The use of thumbnail images, in particular the variance maps, show at least all of the artifacts that were visible in the original data. A long term study to test the stability of DBT systems using the proposed software framework is on-going.

Current approach could be used to enable the automated quality control of DBT systems. The relatively small amount of data per DBT acquisition makes it possible to perform remote QC supervision.

Using the proposed method, performing remotely supervised constancy checking of digital breast tomosynthesis systems becomes a feasible task.

Jacobs, J, Marshall, N, Marchal, G, Bosmans, H, Development and First Experience with a Software Framework for Remotely Supervised Constancy Checking of Digital Breast Tomosynthesis Systems.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8006801.html