5.1. Motivations
Accurate and efficient simulation of x-ray CT scanners plays an important role in research studies.
There are several sources of error and artefacts that affect clinical x-ray CT image quality. The assessment of their effect on image quality is generally commended with the aim to reduce their impact either by optimization of scanners’ design or by devising appropriate image correction and reconstruction strategies. It has long been recognized that the ideal research tool for x-ray CT modelling is the Monte Carlo method. Though, to the best of our knowledge, at the time of this study, there exist only two dedicated MC-based CT simulator which can be used for the purpose of imaging chain modelling. The first is the accelerated Monte Carlo simulator (AMCS) which is a rapid MC-based micro-CT simulator dedicated for modelling of cone-beam small animal x-ray CT scanners [36]. Another MC x-ray CT simulator is CTmod which was used for simulation of single-slice detector configuration and has been used mainly for scatter modelling in the cone-beam geometry [61]. Both simulators have some limitations. In addition, their validation was limited to a specific geometry and the code is not general enough to extend its application to other geometries.
The need to characterize quantitatively the effect of physical and physiological parameters such as contribution of scattered photons, tube voltage, tube current, contrast agent and metallic implants on CT-based attenuation correction in this study spurred the development of an MCNP4C-based x-ray CT Monte Carlo simulator, which allows simulating fan- and cone-beam CT scanner geometries with single-slice, multi-slice and flat-panel detector configurations. Detailed information about the development and validation of our MCNP4C-based x-ray CT simulator is presented in Paper III.
5.2. Development strategy
To use the advantages offered by extensively tested state-of-the-art general-purpose Monte Carlo codes in terms of versatility, published reports and long-term technical support and maintenance, our developed x-ray CT simulator was built on top of the MCNP4C general-purpose MC computer code, which serves as a core layer giving the opportunity to the developer to construct application-specific modules in a hierarchical architecture. Detailed MC radiation transport into the x-ray tube, bow-tie filter, collimator, phantom and detector is performed using MCNP4C with appropriate adjustment of physics treatment options implemented into the code.
In the first step of using MCNP4C code, the user should create an input file which contains information about the problem such as geometry specification, description of materials, type of answer or tally and variance reduction techniques to be used. The creation of input file in the
MCNP4C code using first and second degree surfaces is not an easy task, especially when dealing with the complex geometries typical of an x-ray CT scanner requiring an extremely large number of planes to create detector cells and septa. To solve this problem, a user interface program running under Matlab 6.5.1 (The MathWorks Inc., Natick, MA, USA) was developed by an easy to use concept. Basically, the user is asked to choose the x-ray CT scanner’s design parameters.
According to the information provided by the user, the interface program creates the scanner geometry as input file for MCNP4C. Since MCNP4C is not able to simulate gantry rotation, the geometry of each view is created in separate files and run sequentially [37]. After simulation of all views, the sinogram is created from detector outputs after blank scan correction in all views by the user interface program. The filtered back-projection reconstruction algorithm [122] has been implemented in the user interface program for image reconstruction of simulated data sets. Many other quantities such as absorbed dose, individual scatter and primary profiles can be also extracted from the created output file for each view. Figure 10 illustrate the principles and main components of the MCNP4C-based simulator as applied to model an x-ray CT imaging system.
Fig. 10. Principles and main components of MCNP4C-based Monte Carlo program dedicated for simulation of x-ray CT imaging systems.
5.3. Experimental validation
Generally, before Monte Carlo simulated data can be trusted, the simulation model must be validated. According to the definition of the international standard on quality management systems (ISO 9000), validation is “the confirmation through the provision of objective evidence that the requirements for a specific intended use or application have been fulfilled”. Therefore different intended uses require different validation studies. We will first discuss the intended use of the MC model which underlines this validation study. In the validation study described in this work (Paper III), the intended use of the MC model is the assessment and quantification of the impact of different physical parameters on x-ray CT image quality and its propagation to CT-based attenuation correction in PET. The validity of the MCNP4C-based Monte Carlo simulator was verified by comparing the simulated and measured distributions from various uniform and non-uniform phantoms on both fan- and cone-beam x-ray CT scanners. The single-slice GE HiSpeed X/iF (General Electric Healthcare Technologies, Waukesha, WI, USA) fan-beam CT scanner and the cone-beam small-animal SkyScan 1076 (SkyScan, Aartselaar, Belgium) with flat-panel detector were used for experimental measurements. Good agreement between the simulated and measured projections and reconstructed images was observed.
After the experimental validation study, we concluded that the developed x-ray CT simulator is a powerful tool for evaluating the effect of physical, geometrical and other design parameters on
the performance of new generation CT scanners and image quality in addition to offering a versatile tool for optimizing the absorbed dose to patients and investigating potential artefacts and optimal correction schemes when using CT-based attenuation correction on dual-modality PET/CT units [5] in connection with ongoing research in our lab related to PET quantification using a dedicated PET Monte Carlo simulator [41].
5.4. Applications of the MC x-ray CT simulator
Today’s applications of MC techniques in the field of x-ray CT imaging include performance assessment and optimization of design geometries and scanning parameters [37], scatter characterization and rejection strategies [38, 61], detector configuration and material [87], generation of data sets for testing reconstruction and beam hardening correction algorithms [123, 124] and absorbed dose calculations to assess radiobiological risk from CT scans. For the latter, the accuracy of MC simulations is well established for both axial [115] and spiral [125] scanning modes.
With the advent of multiple-row and flat-panel detector configurations in addition to the slip-ring technology, there have been rapid developments in the design of clinical and small-animal CT scanners including: x-ray tube specifications, geometrical magnification, detector configuration and dose management. MC simulations offer many advantages including the possibility of optimizing tube design, development of new target/filter combinations and calculation of off-axis spectra to improve image quality and reduce patient dose [78]. The optimization of geometrical magnification in x-ray CT, which depends on the source to detector and iso-center distance, is another application of Monte Carlo CT simulators. The optimal detector’s element material and size, which depends on the balance between image resolution, patient dose and signal to noise ratio is another active research area where MC modelling plays an important role.
The corruption of projection data in x-ray CT with scattered radiation decreases low contrast detectability, reduces CT numbers and introduces cupping and streak artefacts in reconstructed images. Scatter removal is also mandatory in x-ray CT imaging because of the need to have clinically acceptable low contrast detectability. The assessment of the scatter component in fan- and cone-beam x-ray CT scanners is an active research area in quantitative imaging and there are many relevant contributions to this domain [38, 61, 81]. The most common technique used to reduce the detection of scattered radiation consists of using collimator plates inside the detector housing (septa) in multi-slice CT and antiscatter grids in flat-panel cone-beam CT scanners. The optimization of septa length and thickness (namely geometrical efficiency of detection system) as well as septa material is also being investigated through assessment of resulting scatter-to-primary ratio (SPR) using MC calculations. It has been shown that increasing the septa length could effectively reduce the contribution of scattered radiation, thus decreasing the SPR at the expense of more manufacturing constraint to avoid possible septa plates’ vibration during the gantry rotation [37].
Analytical models for generation of transmission x-ray projections for testing image reconstruction and beam hardening correction algorithms can be used with some confidence for simple geometries and homogeneous objects; however, their application to more complex geometries and nonhomogeneous objects is complicated and prone to error. A more general and accurate approach for generation of data sets is to use MC simulations by paying special attention to the number of simulated events to reduce statistical uncertainties in the generated data sets.