Congress:
EuroSafe Imaging 2020
Keywords:
Not applicable, Dosimetric comparison, Dosimetry, Cone beam CT, Radioprotection / Radiation dose, Interventional vascular, Interventional non-vascular, Action 9 - Facilitation of research in advanced topics of radiation protection
Authors:
T. Russ, A. M. Abdelrehim, D. F. Bauer, S. Hatamikia, L. R. Schad, F. Zöllner, K. Chung
DOI:
10.26044/esi2020/ESI-05719
Description of activity and work performed
We consider an imaging scenario of a robot-assisted liver-biopsy. In this case, the location of a metallic biopsy needle needs to be determined, which is positioned by the robot. A circular acquisition orbit is not possible in this situation due to the assistance robot effectively blocking the necessary actuation space. Alternatively, a tilted circular (20°-angle) as well as a triple-arc-orbit (180°-40°-180°) are investigated; both are depicted in Figure 2 along with a conventional circular trajectory. The tilted orbit satisfies the Tuy condition and therefore promises solid image quality [2], whereas the triple-arc-orbit provides superior practicability in terms of acquisition time. This is due to the time-consuming step-and-shoot implementation of the tilted trajectory, while the triple-arc can be realized using three dynamic default movements.
For the image quality evaluations, a digital phantom of a water cylinder containing a slanted tungsten wire was generated; the size and orientation of the water cylinder mimics the imaging scenario of a human head. The projection images of the phantom were simulated using GATE, a Monte Carlo simulation toolbox for medical applications based on the Geant4 code. The simulation process incorporates an Artis Zeego detector model, a lead collimator effectively generating the rectangular beam shape, the relevant physical interactions and a realistic signal digitization procedure. The acquisition orbits consisted of 200 projections, each simulated using 2 billion primary photons sampled from a 100 kVp X-ray spectrum. Due to the high computational burden, GATE simulations were paralleled and executed using GateLab, which utilizes CPU clusters of the European Grid Infrastructure [3]. Consequently, the simulation time was reduced from a few weeks down to 3 days.
As an alternative approach, forward projections are obtained using raytracing on a GPU via the ASTRA toolbox for MATLAB [4,5]. The 200 projections of the triple-arc-orbit generated with the raytracing technique are animated in Figure 3. A side-by-side comparison of the GATE and the raytracing projections is shown in Figure 4. While the shape of the cylinder is apparent in both methods, the tungsten wire is only visible in the raytracing projection. This indicates an insufficient photon count in the GATE simulations. In the ideal imaging scenario, the intensity of photons per pixel would amount to several 10³ counts behind the imaged object, whereas in the simulated projections the count is only slightly below 10³.
After simulation, the raytracing projection data was reconstructed by means of the Simultaneous Iterative Reconstruction Technique (SIRT) with 100 iterations on a GPU using the ASTRA toolbox. The subsequent image quality evaluation is performed using an oversampled line-spread-function (LSF) obtained from the slanted tungsten wire for an accurate determination of the modulation transfer function (MTF) [6]. The three resulting MTF curves for the x-dimension are plotted in Figure 5. The MTF of the tilted circular orbit has a mean percentage error (MPE) of only 5.93% compared to the most favorable MTF of the standard circular orbit. At an MPE of 43.73%, the triple-arc-orbit shows a substantial deterioration in spatial resolution. Our evaluations underline the fact that a tilted circular orbit can provide an alternative solution to the conventional trajectory whilst maintaining comparable image quality. If sacrificing nearly half of the spatial resolution is an option, the investigated triple-arc-orbit provides a time-efficient alternative.
For the estimation of physical dose to the patient, we include the XCAT phantom in the imaging scenario [7]. For this task, GATE provides the feature to monitor the deposited energy in each phantom voxel. After converting the Hounsfield units (HU) of the XCAT phantom to physical density values by means of the stoichiometric calibration [8], dose values were calculated from the deposited energy values divided by the mass in each voxel. A sagittal XCAT slice overlaid with a relative dose map is depicted in Figure 6. The field-of-view (FOV) of a CBCT resulting from a circular orbit is clearly visible as a distinct conical dose distribution in the central torso region. Boney structures show more pronounced dose values due the Z-dependency of the photoelectric effect in the diagnostic photon energy range.