Purpose
The new radiation protection laws and rules in Europe and especially in Germany require a more detailed automatic assessment of the applied X-ray dose and image quality of e.g.computed tomography (CT) images. While the applied doses are documented, there is currently no standardized approach for assessing and documenting the image quality. This is also due to the fact that there are large inter-observer discrepancies in evaluating the quality of an image or of a certain region-of-interest. The future goal is to overcome the
limitation of...
Methods and materials
Thorax CT images were collected from more than 450 cases within a European project (MEDIRAD). The long-term goal within the MEDIRAD project is to correlate subjective image quality assessment results from observer studies with objective image quality descriptors. Objective image quality assessment was performed using two approaches:
1) evaluation of the modulation transfer function (MTF) on structures in the region-of-interest and
2) the noise power spectrum (NPS) in that same region.
The area under the MTF curve was integrated and used as an image quality...
Results
Preliminiary results show that thorax CT images labeled with a lower quality tend to have a larger noise variance (estimated from the NPS). Additionaly, those images have a lower MTF.
NPS results (see. Fig. 3) show that the high frequency noise increases with smaller slice thickness. This is also the case for CT images with lower tube currents, e.g. at 10mA.
The obtained results show that an automated image quality analysis of clinical thorax CT images for documentation as well as dose reduction purposes is...
Conclusion
MTF and NPS were estimated in close proximity to structures used for diagnostic purposes. One challenge with computing the NPS is to identify an ROI which contains a homogenous background, i.e. no anatomical structures.
MTF and NPS parameters will be estimated from a labelled thorax CT databases in the future in order to provide an automated method for providing an objective metric for CT image quality.
Personal information and conflict of interest
Zahra Passand; Magdeburg/DE, Email:
[email protected]
Christoph Hoeschen; Magdeburg/DE
References
[None]
Acknowledgements
The work was funded by the European Union in context of the MEDIRAD project (Grant number: 755523).
The authors would like to thank the Bundesamt für Strahlenschutz for their technical feedback and for providing a phantom for the CT measurements.