Gregor Pahn DIPLPHYS, Presenter: Nothing to Disclose

Stephan Skornitzke, Abstract Co-Author: Nothing to Disclose

Jens Hansen DIPLPHYS, Abstract Co-Author: Nothing to Disclose

Hans-Ulrich Kauczor MD, Abstract Co-Author: Research Grant, Boehringer Ingelheim GmbH Research Grant, Siemens AG Research Grant, Bayer AG Speakers Bureau, Boehringer Ingelheim GmbH Speakers Bureau, Siemens AG Speakers Bureau, Novartis AG

Wolfram Stiller PhD, DIPLPHYS, Abstract Co-Author: Nothing to Disclose

Clinical applications of computed tomography (CT) show a trend towards quantitative imaging, offering objective information to aid clinicians in their decisions. In order to guarantee reliable and reproducible results, demands on accuracy and stability of imaging systems are more rigorous than ever, making standardized procedures for quality assurance (QA) imperative. In this study, a novel analysis tool for semi-automatic standardized quantitative evaluation of CT image quality (IQ) is presented and assessed based on a standardized set of phantom measurements.

Using DICOM CT image data, the IQ software developed in-house allows for analysis of CT number accuracy, noise magnitude and power spectrum (NPS), uniformity across the field-of-view, contrast-to-noise ratio (CNR), and spatial resolution (SR) from bar patterns and edges. Maximum information is used by automatically including all slices, enabling quantitation and correction of phantom alignment inaccuracies. The object-oriented software implemented in C++ features a modular, extendable design for maximum automation of IQ assessment and offers a graphical user interface (GUI). Image data of two commercially available phantoms acquired with various protocols on a CT system before and after a detector upgrade is used for evaluating the software’s capabilities.

The software runs stable independent of CT data source. Customizable regions-of-interest (ROI) are automatically placed and propagated throughout all image slices of each IQ phantom section. CT number and noise distributions are automatically filled to 2D and 3D histograms; mean values as well as resp. standard deviations are calculated and also used for CNR evaluation. The results can be fitted algorithmically. Pixel-by-pixel analysis of ROIs yields NPS for noise frequency and edge-spread and modulation-transfer functions for SR assessment. The software allows for comprehensive graphical presentation and data export of all results.

The CT IQ analysis software enables standardized and highly automated IQ assessment. Results are reproducible and reliable due to high statistics. Additional phantom types and evaluation algorithms can easily be implemented.

The novel CT IQ analysis software enables time-effective standardized automated QA and assurance of IQ equivalence for different scanner models and acquisition protocols, e.g. for multicenter studies.

Pahn, G, Skornitzke, S, Hansen, J, Kauczor, H, Stiller, W, Putting Numbers to Images: Comprehensive Quantitative Image Quality Assessment for Computed Tomography Using Semi-automatic Analysis Software.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14016997.html