University of Heidelberg
Faculty of Medicine Mannheim
University Hospital Mannheim
These pages are still under constructions and will be available soon! Please check again later!

If you have questions concerning a specific publication please use this form with subject 'information about publications' and giving the full citation in the message body.

Collaborate Research Projects
Home > Publications > Abstract >

Fully-automated quality assurance in multi-center studies using MRI phantom measurements

M. Davids, F. Zöllner, M. Ruttorf, F. Nees, H. Flor, G. Schumann and L. Schad and the IMAGEN Consortium

Magn Reson Imaging, 32, pp.771-780

Phantom measurements allow for investigating the overall quality characteristics of an MRI scanner. Especially within multicenter studies, these characteristics ensure the comparability of the results across different sites, in addition to the performance stability of a single scanner over time. This comparability requires consistent phantoms, sequence protocols, and quality assurance criteria. Within the scope of this work, a software library was implemented for fully-automated determination of important quality characteristics (comprising signal-to-noise ratio, image uniformity, ghosting artifacts, chemical shift and spatial resolution and linearity) including methods for data preparation, automated pre- and postprocessing as well as visualization and interpretation. All methods were evaluated using both synthetic images with predefined distortions and a set of 44 real phantom measurements involving eight sites and three manufacturers. Using the synthetic phantom images, predefined levels of distortion that were incorporated artificially were correctly detected by the automated routines with no more than 2.6% of relative error. In addition, the methods were applied to real phantom measurements - all data sets could be evaluated automatically considering all quality parameters as long as the acquisition protocols are followed. Shortcomings of the processability only occurred in the ghosting artifacts (39/44 evaluable) and the spatial linearity (43/44 evaluable) analysis due to gross misalignments of the phantom during image acquisition. Based on evaluation results, the accuracy of the evaluation appears to be robust to misalignments, artifacts, and distortions affecting the images, allowing for objective fully-automated evaluation and interpretation of large data set numbers.

Contact: Dr. Frank Zöllner last modified: 21.09.2020
to top of page