Abstract： Objective To investigate the accuracy of image registration with computed tomograpby (CT) and magnetic resonance imaging(MRI) and to determine gross tumor volume(GTV) of hepatocellular carcinoma(HCC).Methods Thirteen patients were enrolled in this study.CT image was taken in inhala tion phase,and MRIT2 image at the time of peak exhalation.Additional MRIT2 image was taken by fast scanning(MRIT2F) in peak inhalation phase in 6 patients.After mutual information method to CT/MRI im age registration,manual adjustment was made to optimize the accuracy of image fusion.The GTV and liver of each patient was independently contoured by two observers on CT,MRIT2 and MRIT2F images.The accura oy of image fusion was assessed by the ratio of liver overlap(P-LIVERCT-MRIAT2 ,P-LIVERCT-MRIT2F) ,and the dis tance between bone markers(DCT-MRIT2,DCT-MRIT2F) of CT and MRI on the fused image.The volumes of GTV contoured on CT (V-GTVCT),MRIT2 (V-GTVMRIT2),MRIT2F (V-GTVMRIT2F) and their overlap (V GTVCT-MRIT2,V-GTVCT-MRIT2F) and composite volumes (V-GTVCT+MRIT2,V-GTVCT+MRIT2F)were measured.The percentage of V-GTVCT and V-GTVMRIT2 on V-GTVCT+MRIT2,V-GTVCT and V-GTVMRIT2F on V-GTVCT+MRIT2F,V GTVCTMRIT2 and V-GTVCT-MRIT2F on V-GTVCT was also calculated,respectively.Results The mean DCT-MRIT2 and DCT-MRIT2F were 2.7±0.8 mm and 2.1Q±0.9mm.The mean P-LIVERCT-MRIT2 and P-LIVERCT-MRIT2F were 85.9% ± 4.1% and 92.7%± 1.5%.Interobserver difference was significant for GIN defined by CT,but not by MRIT2.V-GTVCT,V-GTVMRIT2 and V-GTVMRIT2F were 387±396 cm3 ,488 ±461 cm3 and 597 ±541 cm3 ,respectively.The percentage of V-GTVCT and V-GTVMRIT2 on V-GTVCT+MRIT2 was 66.2%±13.5% and 88.7% ± 10.2% ,while V-GTVCT and V-GTVMRIT2F on V-GTVCT-MRIT2F was 71.3%±12.7% and 93.5%± 4.8%,respectively.Conclusions CT and MRI for image fusion should be obtained in the same respirato ry phase and in the same treatment position.Automatic registration using mutual information method by auto matic registration software is useful.CT and MRI image should be integrated for HCC GTV delineation.GTV used for planning should be the sum of CT-defined GTV and MRl-defined GTV.