In this study,
we evaluate the effect of different levels of iterative CT image reconstruction (IR) algorithms and convolution kernels on the quality of dynamic CT data for the quantitative estimation of myocardial blood flow (MBF).
In the daily clinical setting,
there is increasing awareness for the need to determine MBF quantitatively in ml/100g/min.
This highly important role of absolute MBF estimation deduces from increasing demands for therapeutic interventions in cardiovascular diseases.
MBF estimation is important not only in the evaluation of regional and global perfusion changes caused by coronary stenotic lesions or artheriosclerosis,
but also in quantitatively validating the effectiveness of applied interventional,
surgical or drug therapies.(Makarenko,
et al.
2013)
This clinical importance is accompanied by large improvements of imaging technologies and computing power which enables ameliorations in image reconstruction modalities.
Image Reconstruction
The so called filtered back projection (FBP) reconstruction algorithm has been the standard for CT image reconstruction for many years.
The advantage of this method is the comparatively low computational cost.
However,
in case of noisy or poorly sampled CT projection data,
the result of the images is not satisfiable.(Mehta,
et al.
2013) Nevertheless,
continued progress in CT based diagnosis includes developing scan protocols that reduce the risk of associated radiation exposure.
Since lowering the radiation dose leads to a large increase of the variance in the photon count statistics and thus noisier images,
new reconstruction algorithms are necessary.
Iterative reconstruction (IR) algorithms are based on statistical optimization principles and try and find the "best fitting" image to the acquired projection data.
While FBP equally weights the contributions of all measurements,
in IR,
the noisiest measurements are given lower weights in the iterative process.
This approach results in a reduction of noise and artifacts.
The method not only improves the visual appearance of the image,
but also enables quantitative analysis of the CT data.(Whiting,
et al.
2006)
These advantages of IR lead to an increasing use of such approaches in CT imaging to enable dose minimization.
In many studies,
the quality of the reconstructed images is mainly assessed through visual inspection by radiologists (see e.g.
(Schabel,
et al.
2013)) which is of utmost importance in the daily clinical practice.
However,
one needs to keep in mind that the human eye can distinguish a few gray values only and visual appearance of the image depends on the chosen window level and window width.
The underlying quantitative CT data remain the same as can be seen in the example below (Fig.
1).
Fig. 1: Comparison of different window levels and window widths. The images appear completely different, the quantitative data remain the same.
Thus,
for a quantitative analysis of image data such as perfusion measurements or histogram analysis,
there is a strong need to also quantitatively determine the image quality.
This study assesses changes in HU values for different levels of IR compared to filtered back projection (FBP) in simulated perfusion studies and evaluates the effect on regional perfusion determination.