PURPOSE: To compare the CT findings of hepatocellular carcinomas (HCCs) immediately after radiofrequency (RF) ablation with those obtained at six months' follow-up, to analyse the relationship between degree of shrinkage of a necrotic lesion and marginal recurrence, and to evaluate the factors influencing shrinkage. MATERIALS AND METHODS: We retrospectively evaluated 54 patients with 62 HCCs who underwent only RF ablation between May 1999 and July 2000. For six months after ablation, all had been free from marginal recurrence ad new-growth tumors. The findings of six-month follow-up CT were compared with those obtained immediately after RF ablation, and the volume of each necrotic lesion was calculated and compared. In terms of degree of shrinkage, tumors were classified as belonging to either group I (below 50%), group II (50-80%) or group III (above 80%). Each tumor was analysed in terms of its Child-Pugh classification, vascularity at CT, size, treatment details, the post-ablation appearance of its margins, and the presence, during necrosis, of peritumoral vessles and air bubbles. For statistical evaluation, Fisher's exact test was used. Shrinkage after ablationwas correlated to marginal recurrence during a period of more than one year. RESULTS: The Follow-up CT at six months showed that since immediately after ablation, necrotic lesions had shrunk by an average of 72%. Peritumoral vessels were seen in 12 of 14 cases in group I (86%), three of 13 in cases of group II (23%), and 11 of 35 in group III (31%)(p=0.001). Immediate CT revealed the presence of air bubbles in two cases in group I (14%), five in group II (38%), and 24 in group III (69%)(p<0.05). At follow-up CT performed during a period of more than one year, marginal recurrence was noted in four cases in group I, two in group II and two in group III (p<0.05). CONCLUSION: When a necrotic lesion contains no peritumoral vessels but does contain air bubbles, the degree of post-RFablation shrinkage increases. The greater the shrinkage, the less the marginal recurrence rate.