Session Item

Proton therapy is promising in the treatment of abdominal paediatric malignancies due to its potential in reducing treatment side effects. However, inter-fractional variation in gastrointestinal (GI) gas volume and position may lead to under/over-shooting, compromising target coverage. We investigated the use of surface imaging in paediatric settings to monitor interfractional anatomical variation, by correlating internal variation in GI gas volume with simulated external metrics of surface change.

Data from 21 patients treated with IMAT for abdominal high-risk neuroblastoma (median age 4, range 2-19 y) were used in this study. The body and GI gas volumes were semi-automatically delineated on 21 CT and 77 weekly CBCT scans using ITK-Snap. The CBCTs were rigidly co-registered to the CT using NiftyReg, and both scans translated to treatment isocenter. CBCTs were considered as treatment position, while the CT was the reference position. Quantitative metrics of internal and surface changes were calculated for each pair of CT/CBCT scans within the common imaging field-of-view, as defined in Fig.1. Surface parameters were simulated from the body contour. The anterior surface was extracted from the body and converted to a point cloud. Point clouds were registered using the iterative closest point algorithm in MATLAB to estimate the translational and rotational corrections needed to align the CBCT surface to the reference. These metrics are surrogates to the correction that would have been obtained with surface imaging in treatment position. GI gas variation was then correlated with all metrics. Statistical analysis was performed in Stata 16.1 (5% significance level).

We found a ΔV_gas, Δd_body and Δd_surface of -75.5±113.7ml, -1.1±1.5mm and -2.6±3.3mm, respectively, when pooling all data for analyses. There was a trend of reduction in GI gas and body contour from planning. Separation of body contour and surface correlated moderately to strongly with GI gas (Fig.2). A stronger correlation was found between ΔV_gas and Δd_surface (R=0.58) than between ΔV_gas and Δd_body (R=0.46). Regarding metrics of surface correction, the strongest correlation with ΔV_gas was found for t_Y (R=0.58) and r_x (R=-0.35), anterior-posterior translation and rotation of the left-right axis, respectively. These findings suggest that anterior surface changes were likely more affected by abdominal distension caused by GI gas, while body contour changes were possibly more affected by other inter-fractional variations such as weight fluctuation and setup errors.

Interfractional variation in GI gas volume correlated well with surface variation metrics. Surface imaging may be useful to inform paediatric image guidance protocols and reduce the need for CBCT and associated dose burden, by predicting timepoints with internal changes that warrant additional volumetric imaging. Further work is required to validate our findings using real instead of simulated surface imaging data.