Safety margin assessment after radiofrequency ablation of the liver using registration of preprocedure and postprocedure CT images

The purpose of our study was to evaluate a new technique for registration of postprocedure to preprocedure CT images to determine the effect on the safety margin assessment after radiofrequency ablation (RFA). Registration of post-RFA CT to pre-RFA CT images was performed using prototype software via nonrigid registration in 31 patients with hepato-cellular carcinoma who were treated with RFA. Registration accuracy was validated by setting pairs of corresponding landmarks on registered post-RFA CT and pre-RFA CT images and by calculating the mean difference between the corresponding landmarks. Three radiologists independently conducted a retrospective review of the pre-RFA and post-RFA CT images for safety margin assessment with and without registered images. The safety margin was rated using a 4-point scale (1, residual tumor; 2, < 2 mm safety margin; 3, 2 to < 5 mm safety margin; and 4, ≥ 5 mm safety margin). Interobserver agreement was evaluated using the weighted kappa test. The mean difference between the corresponding landmarks was 1.3 mm, thus suggesting accurate registration. A more accurate correlation between the degree of the reader safety margin and the reference value was obtained from pre-RFA and post-RFA CT images using registered CT images than without using registered images (γ, 0.918 vs 0.887 for reader 1, 0.888 vs 0.651 for reader 2, and 0.811 vs 0.497 for reader 3, respectively). Interobserver agreement (kappa) increased from 0.503-0.558 in the first session without registered images to 0.807-0.869 in the second session with registered CT images. Registration of post-RFA CT to pre-RFA CT images is an accurate and useful technique for assessing the safety margin immediately after RFA.