
Discussion
The phantom measurements point out that the displacement could be determined from the oblique radiation beams of the Tetrahedron technique with the same accuracy as from the orthogonal verification beams. The differences between measured displacement and setup displacement can completely be attributed to the accuracy of the table movement (0.1 mm precision), gantry rotation (1°) and the image resolution (0.4 mm per pixel).
However, this result only holds for phantoms and not for real patients. In the case of patient images the situation becomes much more complex. First, the interpretation of the images is more difficult. The contrast of the bony structures in the portal images is low and the images of the verification beams of only 4 MU are relatively noisy. Secondly, the head and neck region of a patient is not a rigid body. Therefore, also rotations and deformation can occur. In the portal images rotations can not always be distinguished from translations which lead to errors in the determination of the setup error.
The problem with the interpretation of patient data is that the real setup error is not known. It is current practice to determine the patient setup error from portal images of an AP and a lateral verification beam. Therefore, we have compared the setup errors from the oblique radiation beams with those determined from the AP and lateral beams. However, one should bear in mind that also the setup error determined from the AP and lateral beams contains measurement errors.
Bel et al. have investigated whether a wedged pair of oblique beams could be used for the position verification of patients with parotid gland or tonsillar tumors2. They concluded that it was not possible to obtain consistent translational setup deviations using bony structures, due to patient rotations.
The results of the patients treated with the Bellinzona technique show a good agreement between the setup errors determined from the oblique radiation beams and from the orthogonal verification beams in the ventrodorsal and craniocaudal direction (Table 2). The inaccuracies in the setup errors can be explained by the inaccuracies in the gantry angle (1°), in the review process of the portal images and in the calculation of the setup errors from the oblique fields. The inter and intraobserver variability in the review of patient images is about 1 mm (1 SD) in our institute (data not shown). Therefore, we assumed an accuracy of 1.0 mm (1 SD) in the setup errors determined from the AP and lateral portal images. We assume a normal distribution of the setup errors. This indicates an accuracy of 1.7 mm (1 SD) in the setup errors determined from the oblique radiation beams in the lateral and 0.8 mm (1 SD) in the ventrodorsal and craniocaudal directions.
The accuracy of the calculation of the setup errors from the oblique fields depends on both de difference in gantry angle between the oblique fields6 and the distance to the isocenter of the bony structures used in the image analysis4. Based on the results of KolkmanDeurloo et al. it can be expected that the maximum deviations in the calculation for the Bellinzona technique (with a difference in gantry angle of about 40 °) will be twice that of a calculation with a difference in gantry angle of 90°.
The difference between the oblique radiation and the verification beams has decreased in the lateral and in the craniocaudal direction by adding an AP verification beam to the oblique beams (Table 2). The difference in the ventrodorsal direction was not influenced by the addition of the AP beam. In this situation the AP verification beam is used in the determination of the setup errors in both situations. Hence, the measurements are not independent anymore and a quantitative interpretation of these data is not possible. However, we assume that the lateral setup error can be obtained with the same accuracy from the combination of the AP and oblique beams as from the AP and lateral beams. The accuracy of the setup error in the ventrodorsal and craniocaudal direction determined from the combination of the oblique and AP beams is about 1 mm. Because the information content of the portal images (contrast, amount of bony structures, beam of view) of the oblique and orthogonal beams is comparable, this result is in agreement with what one would expect.
For the Tetrahedron technique with two oblique beams the accuracy in the setup errors from the oblique beams will be 1.4 mm (1 SD) in the lateral and the craniocaudal direction and 0.7 mm in the ventrodorsal direction. If three oblique beams are used the accuracies (1 SD) are 1.2 mm in the lateral, 0.8 mm in the craniocaudal direction and 0.5 mm in the ventrodorsal direction. The accuracy in the ventrodorsal direction is much better than 1.0 mm. Only relatively small setup errors were found in the ventrodorsal direction. However, the difference in setup errors does not depend on the magnitude of the setup error. Therefore, the most probable explanation is that the accuracy in the review process in the ventrodorsal direction is better than 1.0 mm (1 SD).
It is expected that the accuracy in the review of the radiation beams is less than that of the verification beams because the beam size and therefore the field of view is smaller and as a result less bony structures can be recognized and the interpretation of the images is harder. The results for the lateral and craniocaudal setup errors are in agreement with this assumption. Especially if only two radiation beams are used the accuracy in the setup errors determined from these radiation beams is less than that of the verification beams. In the ventrodorsal direction the accuracy of both methods seems to be comparable.