Abstract： Objective To compare the accuracy and feasibility among CT, MR, and CT-MR image fusion technology in the delineation of gross target volume (GTV) for nasopharyngeal carcinoma (NPC). Methods Thirty-six consecutive patients with newly diagnosed or recurrent NPC were enrolled. Each pa-tient underwent both CT and MR scanning in the same treatment position. Several lead marks were made at the exactly same locations of the body surface before CT and MR scanning. The two sets of images were then transferred to the Tomcon workstation for image fusion. CT-MR image registration was performed using Land-Mark methods. GTV of each patient was contoured on CT (GTV_(CT)), MR (GTV_(MR)) and CT-MR (GTV_(CT-MR)) images. Results The mean GTV_(CT), GTV_(MR) and GTV_(CT-MR) were 27.60 cm~3, 30.99 cm~3 and 31.71 cm~3, respectively (F = 7.48, P = 0.001). Significant difference was found between GTV_(CT) and GTV_(MR) (q =2.54,P=0.016), GTV_(CT) and GTV_(CT-MR) (q =3.10,P =0.004), but not GTV_(MR) and GTV_(CT-MR) (q = 1.31 ,P = 0. 199). Significant difference among GTV_(CT), GTV_(MR) and GTV_(CT-MR) was found in patients with skull base invasion (35.65, 42.70 and 44.22 cm~3 ; F = 14. 13, P = 0. 000), but not in those without skull base invasion (20.79 cm~3, 20.46 cm~3 and 21.18 cm~3 ; F = 0.18, P = 0.832). Significant difference between GTV_(CT) and GTV_(CT-MR) was found in patients with T_3 and T_4 tumor (t = -2.17,P =0. 036), but not in those with T_1 and T_2 (t = -0.66 ,P = 0.514). Conclusions CT-MR image fusion is helpful in GTV de-lineation for NPC, particularly for patients with skull base invasion.