With the pelvic ROI demonstrated here with Ellen (but including less legs from caudal parts, in fact cutting the additional legs extensions outside of the ROI, leading to rather narrow ROI in CC direction), pitch in the kV-LAT images were evaluated from 45 AlignRT (A) patients. For comparison, 45 laser setup (L) patients were evaluated. In total, 1850 images were offline evaluated (969 L, 881 A). Errors between symphysis and L5 in AP direction (VRT), demonstrating pitch, in LAT images exceeded 5 mm (A) in 7% and (L) in 11% of the fractions. Corresponding exceeding between L5 and symphysis were in LAT direction (A) 9% and 10% (L) in AP images, demonstrating the rotation. These all values given, show the errors in the images between the structures after setup, before IGRT and online match corrections. We think that pelvis has been one of the easiest site to setup with AlignRT, and accurate enough. Often posture of the patient is, based on AlignRT, good in the first place and there is no need to touch the patient. So, title of this paper https://doi.org/10.1016/j.tipsro.2021.11.005 based on my experience seems to fit with pelvis setups as well. Can AlignRT be significantly more accurate positioning method than laser setup in pelvis in general from rotations part? May be, but have been satisfied with our accuracy with simple and quick workflow and decided not to fine-tune it in general, allowing AlignRT deltas at the positioning to be within 3° thresholds from pitch part, for example.
Unfortunately I do not have data from problematic cases concerning pitch, and how those were solved with the aid of AlignRT. Offline image evaluation has been in big role here to correct systematic pitch errors. We teach and ask patient to move a pelvis according to found systematic pitch error in the offline images, to lift a back a little, to press it towards couch… Usually this helps and new reference capture is taken after patient reaches a good posture in the images.