Reproducibility of surface acquisition and setup procedure was better than 0.5mm and
0.5 ? and 1mm and 0.4 ? , respectively. The system accuracy was better than 1mm and 1 ? when a
Sentinel image was used as reference. A global worsening of Sentinel performance was observed
using as reference an external surface extracted from CT study. This effect is probably due to small
differences in considered surfaces, caused by different imaging modalities. The results obtained by
testing the system on rigid phantoms were comparable to those obtained using CBCT and better
than those obtained with conventional portal imaging systems.