Session Item

In ion beam therapy, the dose distribution within a patient is calculated using a planning CT. The obtained Hounsfield units from the CT image must be converted to stopping powers (SP), describing the energy loss per unit path length of an ion within a patient. To avoid range errors in the calculated dose distribution arising from this conversion, the concept of imaging with ions was developed. If the planning CT was measured with the same particle species as the treatment, the image would directly return the SP without any further conversion. However, due to multiple Coulomb scattering, substantial deviations from a straight path for ions in matter are occurring, thus complicating image reconstruction with high precision.

A typical ion CT (iCT) setup consists of two particle trackers, one upstream and one downstream the patient or a phantom and an energy measurement device (calorimeter). With the trackers, the position and direction of each ion are measured in order to estimate its trajectory. The projection value, i.e., the water equivalent path length (WEPL), is based on the determination of the ion’s energy loss along its path by measuring its residual energy in the calorimeter.
A simplified iCT system was modelled in the Monte Carlo simulation toolkit Geant4. As phantoms, two Catphan modules, high resolution and sensitometry, were studied to evaluate SP accuracy and line pair resolution. For image reconstruction, the tomographic iterative GPU-based reconstruction toolbox (TIGRE), initially developed for cone beam CT, was used and adapted to the iCT reconstruction problem.
An existing approach for iCT image reconstruction, based on optimised ion radiographies as input data, was refined and implemented in the TIGRE toolbox as a pre-processing step to image reconstruction. Reconstructions were generated after integrating the proposed additional preprocessing step prior to the ASD-POCS algorithm. Results were compared to a previous study, where TIGRE was used for iCT image reconstruction using a straight-line approach for the ion path only.

The code extension was implemented using the same layered structure as offered by the TIGRE toolbox, i.e., a user-friendly Matlab header while having GPU-accelerated code in the bottom layer to reduce computation time.  Significant improvements regarding SP accuracy (mean absolute percentage error below 1%) and line pair resolution (over 6 lp/cm) were achieved with the new pre-processing compared to the previous study using a straight-line approach.

The TIGRE toolbox was extended to the iCT reconstruction problem by implementing the calculation of optimised ion radiographies. The newly implemented reconstruction workflow was successfully tested with Monte Carlo data.