Session Item

Small animal particle irradiation requires repeatable high geometric and dosimetric precision, especially when multiple irradiation fractions of small structures are required. Often, 3D image guidance is no option at particle beam lines, but 2D X-ray Computed Radiography (CR) may be installed at or close to the irradiation spot. Here, we present a method to reconstruct the full positional/rotational information of a mouse brain from a single 2D CR image via optimization of the mutual information between the image and a maximum intensity projection retrieved from a reference CT. We combined this with a statistical robustness analysis and Monte Carlo-based treatment planning to reduce positional/rotational uncertainties to a negligible level.

We have developed a statistical method of robustness analysis for small-animal irradiation. Statistical distributions of misalignment in all six degrees of freedom of a target region are extracted from a set of CTs via pair-wise 3D-registration. A dose distribution is then calculated for each CT with Monte Carlo and the dosimetric impact of misalignment is determined by calculating statistical distributions of range shift and of DVH curves.
In addition, we have developed a 3D position/orientation reconstruction algorithm that matches a CR image to a synthetic CR image calculated on the fly from a CT image at varying angles (see figure). This is achieved via preconditioned stochastic optimization of the mutual information. We tested the accuracy of the algorithm by comparing the reconstructed 3D position/orientation to a CBCT taken with the mouse in the same position as for the CR.
The presented methods were developed for the irradiation of the hippocampus of mice with a pristine proton Bragg peak with sub-millimeter precision. We tested the methods on CT and CR images from sets of 42 and 10 C3H/He and C57BL/6 mice, respectively.

For our two CT mouse data sets, the robustness analysis revealed statistical misalignments of the mouse skull with standard deviations of 0.64, 0.58, 0.45 mm in x- y- and z- direction and 1.51°, 3.68° and 5.32° for the Euler angles yaw, pitch and roll, respectively. The misalignments in pitch, x- and y- direction are corrected for due to experimental design. Misalignments in z-direction have a negligible dosimetric impact. The misalignment in roll leads to a relevant range-shift of +0.5/-0.7 mm.
Via 3D reconstruction, we could reduce the misalignment to ±1° for all angles and ±0.1 mm in all directions. 3D reconstruction and MC dose calculation with 10^7 particles are achieved in 13 and 2 minutes, respectively (standard scientific workstation with 16 CPU threads).

We have developed an image guidance framework that allows to plan small animal particle irradiation in full 3D precision based on a single X-ray scan via 3D reconstruction and MC dose calculation in about 15 minutes. This enables individual high-precision treatment planning prior to irradiation without the need for a CT.