Objective:To study the effect of rotational errors on the positioning accuracy (PA) and to assess whether correcting rotation in patients with head-neck tumors in radiotherapy or not.Methods:The image information of 34 patients with head-neck tumors treated at Zhongnan Hospital of Wuhan University between August 2019 and January 2020 was collected. Mega-voltage computed tomography (MVCT) images of each patient were taken before radiotherapy, and were registered with planned kilo-voltage computed tomography (KVCT) images by two registration methods. All information was divided into control group (translation only) and intervention group (translation and rotation) according to different registration methods, there were 144 fractioned registered images for each group, respectively. The position errors of the two registration methods were recorded and compared. Data were carried out with Wilcoxon signed rank test and Spearman rank correlation.Results:Translational errors of the control group and the intervention group were 0.10 (5.35) mm and 0.00 (5.78) mm in right-left direction, and there was a statistically significant difference ( Z=-2.675, P=0.007); 0.75 (2.78) mm and 0.60 (2.68) mm in superior-inferior direction, and there was a statistically significant difference ( Z=-2.819, P=0.005); 0.10 (0.90) mm and 0.20 (1.28) mm in anterio-posterior direction, and there was a statistically significant difference ( Z=-3.984, P<0.001). Rotational errors of the intervention group were -0.20 (0.60)°, 0.35 (2.00)°, 0.00 (0.98)° in pitch, roll, yaw, respectively. The distribute of 3D vector corrected frequency for two groups was positively skewed. The corrected cumulative frequency (CCF) varied with 3D vector, 3D vector was 8.0 mm, and 19 F and 16 F fractioned treatments of the control group and the intervention group were not corrected, respectively; 3D vector was between 8.0-13.5 mm, the corrected tendency of the intervention group was slower and fractioned treatment was completed later. The analytical results of Spearman rank correlation showed that rotational errors in pitch were negatively correlated with translational errors of the control group in superior-inferior direction ( r=-0.182, P=0.029) and the intervention group in anterio-osterior direction ( r=-0.484, P<0.001); rotational errors in roll were negatively correlated with translational errors of the intervention group in right-left direction ( r=-0.334, P<0.001); rotational errors in yaw which were positively correlated with translational errors of the intervention group in right-left direction ( r=0.370, P<0.001) were negatively correlated with translational errors of the control group in superior-inferior direction ( r=-0.171, P=0.040) and the same was true for the intervention group ( r=-0.203, P=0.015); total angles were positively correlated and negatively correlated with translational errors of the control group in superior-inferior direction ( r=0.246, P=0.003) and anterio-posterior direction ( r=-0.188, P=0.024), and positively correlated with 3D vector of the control group ( r=0.198, P=0.017), total angles were positively correlated with translational errors of the intervention group in superior-inferior direction ( r=0.170, P=0.041) and with 3D vector of the intervention group ( r=0.239, P=0.004); there were no correlations between rotational errors and the other translational errors (all P>0.05). Conclusion:Although the corrected rotation increases translational errors in anterio-posterior direction and 3D vector, it improves PA for head-neck tumors in radiotherapy. When rotational errors are not corrected, rotational offsets are present with corrected translation to decrease its effect on PA.