PURPOSE To propose a postprocessing technique that measures tumor surface with insufficient ablative margins (≤ 5 mm) on magnetic resonance (MR) imaging to predict local tumor progression (LTP) following radiofrequency (RF) ablation. MATERIALS AND METHODS A diagnostic method is proposed based on measurement of tumor surface with a margin ≤ 5 mm on MR imaging. The postprocessing technique includes fully automatic registration of pre- and post-RF ablation MR imaging, a semiautomatic segmentation of pre-RF ablation tumor and post-RF ablation volume, and a subsequent calculation of the three-dimensional exposed tumor surface area. The ability to use this surface margin ≤ 5 mm to predict local recurrence at 2 years was then tested on 16 patients with cirrhosis who were treated by RF ablation with a margin ≤ 5 mm in 2012: eight with LTP matched according to tumor size and number and α-fetoprotein level versus eight without local recurrence. RESULTS The error of estimated tumor surface with a margin ≤ 5 mm was less than 12%. Results of a log-rank test showed that patients with a tumor surface area > 425 mm(2) had a 2-year LTP rate of 77.5%, compared with 25% for patients with a tumor surface area ≤ 425 mm(2) (P = .018). CONCLUSIONS This proof-of-concept study proposes an accurate and reliable postprocessing technique to estimate tumor surface with insufficient ablative margins, and underscores the potential usefulness of tumor surface with a margin ≤ 5 mm to stratify patients with HCC treated by RF ablation according to their risk of LTP.