Background and purpose: With the development of therapy equipments and technology

the treatment for nasopharyngeal carcinoma(NPC) has entered into the era of precision radiotherapy

and setup errors have become a very important factor affecting treatment effects. The purpose of this study was to analyze the set-up errors detected by the kilovoltage cone beam CT(EPID) and the megavoltage electronic portal imaging device(CBCT) using 2 kinds of different immobilization techniques (pillow + head neck shoulder mask and vacuum bag + head neck shoulder mask) for NPC patients. Methods: A total number of 40 NPC patients were randomly assigned into 2 groups (pillow + head neck shoulder mask group and vacuum bag + neck shoulder mask group). Then each group was further divided into CBCT scan group and EPID group for verification before treatment delivery. We matched the EPID images with the DRRs and acquired the set-up errors in x

y

z axis. Setup errors of CBCT were calculated according to its matched and planned CT images in left-right (x)

superior-inferior (y) and anterior-posterior (z) directions. Paired t-test was used to evaluate the differences. Results: In the pillow + head neck shoulder mask group

the set-up errors of CBCT in the x

y

z axis were x (0.67±2.01)mm

y (0.51±1.71)mm and z (0.57±2.04)mm

respectively. The errors of EPID were x (0.69±2.19)mm

y (0.54±2.03)mm and z (0.61±2.11)mm. In the vacuum bag + head neck shoulder mask group

the set- up errors of CBCT in the x

y

z axis were x (0.42±1.81)mm

y (0.33±1.55)mm and z (0.50±1.75)mm

respectively. The errors of EPID were x (0.44±1.87)mm

y (0.43±1.70)mm and z (0.54±1.77)mm. The vacuum bag + head neck shoulder mask fixed technique was more accurate when compared to the pillow + head neck shoulder mask fixation method (P0.05). Conclusion: CBCT and EPID were similar in detecting set-up errors for the NPC patients. However

the vacuum bag+neck shoulder mask fixed technique was more accurate when compared to the pillow + head neck shoulder mask fixation method.