1231 - Clinical Implementation of Body Surface Monitoring Based on Cone Beam CT for Adaptive Pediatric Proton Therapy

Purpose/Objective(s): Cone Beam CT (CBCT) images are routinely acquired for patient setup purposes but are primarily used qualitatively. We present a CBCT-based monitoring system that quantitatively evaluates interfractional patient body surface changes associated with rapid weight fluctuations resulting from steroid use, concurrent chemotherapy, or anesthesia. An automated tool computes body surface variations along the proton beam path to serve as an early warning system for deviations during the course of radiotherapy, thereby supporting adaptive proton therapy decisions in pediatric patients.

Materials/Methods: A spot-based body surface detection method was developed to compare the planning CT and daily CBCT images. For each proton spot, intersections with the body surface were calculated on both image sets, and surface differences along the beam path were quantified. A single representative metric, defined as the median of all spot-based calculations, was reported for each proton field. The method was validated using a 3D cylindrical water equivalent virtual phantom and a sample proton arc plan comprising 36 beams spaced at 10° increments. A MATLAB-based algorithm was developed as an extension within the clinical image software to automatically detect body surface changes from daily CBCTs exported to the clinical platform. The automated surface change workflow executed in less than one minute without user interaction and recorded the median surface changes for each field between the original CT and CBCT images to a spreadsheet. The method was retrospectively evaluated on pelvis (N=26) and abdomen (N=20) patient cases with a total of 46 patients, including replanned (N=16) and non-replanned (N=30) cohorts, comprising 105 individual proton fields and 889 daily CBCTs.

Results: Phantom validation demonstrated a mean measurement uncertainty of 0.5 mm with a standard deviation (SD) of 0.3 mm across 36 fields when compared with physical ground truth measurements. In pelvis cases, replanned patients (patients = 14, fields = 31, CBCTs = 398) exhibited larger body surface differences, with a mean of 2.3 mm (SD 2.0 mm), compared with non-replanned cases (patients = 12, fields = 30, CBCTs = 250), which showed a mean of 1.3 mm (SD 1.0 mm). Similarly, in abdomen cases, replanned patients (patients = 2, fields = 5, CBCTs = 41) demonstrated greater surface changes, with a mean of 2.7 mm (SD 0.7 mm), while non-replanned cases (patients = 18, fields = 39, CBCTs = 200) showed a mean of 1.4 mm (SD 1.4 mm). These results highlighted the increased mean body surface changes in cases that historically required replanning.

Conclusion: An automated CBCT-based surface detection method was developed to monitor interfractional body surface changes, providing quantitative metrics that enable customized clinical thresholds to trigger dose recalculation for treatment plan quality verification.