Abstract
2363
Objectives Low-dose Cone-Beam CT (CBCT) with a flat panel detector is being used in SPECT/CT imaging to provide localization and attenuation correction. However, effective scatter correction without increasing noise remains a challenging task for low-dose CBCT: Even when using an anti-scatter grid, simulations for patient shoulder regions estimate scatter-to-primary ratios of well beyond 500% at the detector. Directly subtracting so high estimated scatter levels from projections increases noise and may introduce severe streak artifacts. Our work provides a novel method to compensate scatter in low-dose CBCT projections while avoiding streak artifacts.
Methods We compensate scatter fluence with an iterative multiplicative correction followed by adaptive filtration. The iterative multiplicative correction is an efficient approximation of the statistical Poisson MLEM for scatter correction. The adaptive filtering step is the key to suppress noise: It filters each projection pixel with a locally adapted Gaussian low-pass filter, whose width is inversely proportional to the signal intensity. This method is evaluated on phantom and patient studies acquired on the offset-detector geometry flat-panel CBCT of a Philips BrightView XCT system.
Results For a 42cmx27cm elliptical cylinder water phantom, adaptive filtering after scatter correction can reduce the standard deviation of a reconstructed homogenous water section by 20%. In patient studies, soft tissue in the shoulder region has a typical standard deviation of 15HU. Without adaptive filtering, scatter correction introduces streak artifacts and increases the standard deviation to 30HU. By using adaptive filtering, streak artifacts are effectively reduced and the standard deviation is brought back to the same level as without scatter correction with no visible loss of anatomic detail.
Conclusions While conventional scatter correction can introduce strong streak artifacts in low-dose CBCT, the combination with adaptive filtration effectively reduces such artifacts and significantly improves image quality