Recovery of true scatter in blocked regions for blocker-based scatter correction of CBCT - Recovery of true scatter in blocked regions for blocker-based scatter correction of CBCT
Scatter correction (SC) is critically important to mitigate problems of shading artifacts, reduced contrast and
inaccurate CT numbers of con-beam computed tomography (CBCT) for accurate and precise radiation dose
delivery for image-guided adaptive radiotherapy. Among numerous SC methods, the use of lead-strip blockers
is low-cost and easy to implement, where the signal detected in the blocked region is deemed scatter under an
ideal condition to estimate scatter in the unblocked region. A moving-blocker approach has been developed to
estimate scatter and reconstruct the volume image simultaneously from a single CBCT scan and holds the
additional advantages of 1) high accuracy potential with subject-dependent measurements; 2) simple lead strip
design; and 3) substantial dose reduction benefit (~50%). Since the signal in the blocked region is not pure
scatter, heuristic parameter adjustment (denoted as manually tuned scatter estimation, MTSE) has to be done
to achieve satisfying SC performance. Moreover, MTSE is hard to optimize these parameters in a spatially
variant way and could produce heterogeneous reconstruction results, where large CT number errors (>100
HU) in some regions could occur and were clinically unacceptable (> 2% dose error) for treatment planning.
In this work, we propose to overcome the problems of MTSE using rigorous modeling of the blocker-based
image acquisition and deconvolution based scatter estimation (DBSE). Our goal is to eliminate manual
parameter tuning and to achieve acceptable accuracy of reconstructed CT numbers over all regions of interest
for dose calculation (< 100 HU, i.e. < 2% dose error) and significantly improved soft tissue contrast for effective
organ segmentation. This goal will be achieved by the following specific aims: 1) to model the detector
response function and the penumbra effect, which contaminate the scatter signal in the blocked region with the
primary signal in the unblocked region; 2) to develop deconvlution methods for DBSE (frequency-domain
methods, maximum likelihood estimate, and constrained optimization); and 3) to validate the model and to
evaluate DBSE using Monte Carlo simulation and existing data of physical phantoms using clinically relevant
criteria. Successful completion of this research will greatly improve the reliability, robustness, and accuracy of
the moving-blocker technology and other blocker-based methods for CBCT, which is an important step to
translate them into clinic for accurate soft-tissue localization and dose calculation in adaptive radiotherapy.
