Wei Chen MEng, Dipl Eng, Presenter: Nothing to Disclose

Yiannis Kyriakou PhD, Abstract Co-Author: Nothing to Disclose

Willi A. Kalender PhD, Abstract Co-Author: Consultant, Siemens AG Consultant, Bayer AG Founder, CT Imaging GmbH Scientific Advisor, CT Imaging GmbH Shareholder, CT Imaging GmbH Founder, Artemis Imaging GmbH CEO, Artemis Imaging GmbH Shareholder, Artemis Imaging GmbH

The aim of this study was to develop a fast simulation-based scatter correction using a hybrid (GPU/CPU) platform for the reduction of inhomogeneities such as cupping and streaks in CBCT.

The scatter correction algorithm uses an estimate of the scatter intensity based on deterministic and Monte Carlo methods [Kyriakou et al., Phys. Med. Biol. 2006, 51(4567–4586)] to assess single and multiple scatter, separately. For an efficient employment of computational resources, the calculation of single scatter was executed on GPUs, and CPUs were used to concurrently estimate multiple scatter. The scatter distribution was iteratively estimated and subtracted from the projections. To improve the initial volume which was used for the scatter estimation we combined this scatter correction algorithm with an empirical cupping correction (ECC). An volume and detector down-sampling technique and an angular interpolation method were applied for further acceleration. The correction algorithm was evaluated using simulated data for which scatter was generated by Monte Carlo simulations and for head and body datasets using polyenergetic spectra of 80 and 125 kV. The scatter correction quality factor Q (Q=1 for the ideal image, Q=0 for the uncorrected image) was calculated. Experimental validation included measurements of anthropomorphic phantom data on a C-arm FD-CT system (Artis Zeego, Siemens Healthcare, Forchheim, Germany). Evaluations were performed on a 2.6 GHz Intel Xeon Quad core PC, equipped with a NVidia GTX 285 graphics card.

The proposed correction method provided a Q of up to 0.89 for the simulated data by significantly reducing scatter artifacts and increasing contrast. In both simulations and measurements, the number of iterations for the correction was reduced when the initialization was done with the ECC method. The hybrid platform achieved a near real-time performance, requiring below 1 second per iteration. The scatter calculation was accelerated up to 10 times as compared to multi CPUs routine without compromising image quality.

Our simulation-based scatter correction algorithm achieves a real-time correction of cupping and streaks in CBCT images.

The proposed solution has the potential to allow for a new class of advanced scatter correction strategies in clinical practice.

Chen, W, Kyriakou, Y, Kalender, W, Real-time Simulation-based Scatter Correction for Cone-Beam CT (CBCT) Using a Hybrid Hardware GPU/CPU Platform.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9008660.html