Determination of Optimal Icbct Reconstruction Parameters for a New Imaging System on a C-Arm Linac
Abstract
Purpose
To enhance image quality for image-guided radiotherapy (IGRT) by determining the most optimal combination of filter and noise suppression level used in iterative reconstruction of cone beam CT data (iCBCT) acquired with a new imaging panel on a C-arm linac.
Methods
Raw CBCT data from various patient scans was reconstructed with different combinations of filters (Standard, Smooth, Sharp) and noise suppression levels (Very Low, Medium, Very High). Two surveys were designed: scans that used ‘head’ parameters (100 kV, 149 mAs) and scans that used ‘pelvis’ parameters (125 kV, 1093 mAs). Each survey consisted of 20 separate side-by-side comparisons of the different reconstruction combinations. In each comparison, reconstructed images were not labeled and were presented in no particular order. 11 physicists and 4 physicians from our institution participated in this survey, selecting their preferred image quality in each of the 40 comparisons. A 95% two-way cluster bootstrap confidence interval (CI) is used to determine the preferred iCBCT image reconstruction parameters. CatPhan scans were used to evaluate spatial resolution and low contrast resolvability using the same scan parameters and iCBCT reconstruction combinations.
Results
The preferred filter for both scan parameters was Sharp, with a CI of [52.6%,77.4%]. A CatPhan scan reconstructed with each filter resulted in spatial resolutions of 6, 7 and 8 lp/cm respectively. The CI for Medium noise suppression preference was [35.0%,67.2%]. The spatial resolution decreased by 1 lp/cm for each filter when noise suppression increased from medium to high.
Conclusion
To provide the most preferable image quality for IGRT, CBCTs acquired with our new C-arm linac imaging panel should use a Sharp filter and Medium noise suppression for iCBCT reconstruction. This is supported through CI calculation from an intensive survey provided to staff at our clinic, and agrees with brief spatial resolution and low contrast resolvability CatPhan analysis.