[Impact of planning CT slice thickness on the accuracy of automatic target registration using the on-board cone-beam CT].

We have evaluated relationship between planning CT slice thickness and the accuracy of automatic target registration using cone-beam CT (CBCT). Planning CT images were acquired with reconstructed slice thickness of 1, 2, 3, 5, and 10mm for three different phantoms: Penta-Guide phantom, acrylic ball phantom, and pelvic phantom. After correctly placing the phantom at the isocenter using an in-room laser, we purposely displaced it by moving the treatment couch and then obtained CBCT images. Registration between the planning CT and the CBCT was performed using automatic target registration software, and the registration errors were recorded for each planning CT data set with different slice thickness. The respective average and standard deviation of errors for 10 mm slice thickness CT in the lateral, longitudinal, and vertical directions (n=15 data sets) were: 0.7 +/- 0.2mm, 0.8 +/- 0.2mm, and 0.2 +/- 0.2mm for the Penta-Guide phantom; 0.5 +/- 0.4 mm, 0.6 +/- 0.3 mm, and 0.4 +/- 0.3 mm for the acrylic ball phantom; and 0.6 +/- 0.2 mm, 0.9 +/- 0.2 mm, and 0.2 +/- 0.2 mm for the pelvic phantom. We found that the mean registration errors were always less than 1 mm regardless of the slice thickness tested. The results suggest that there is no obvious correlation between the planning CT slice thickness and the registration errors.