OBJECTIVE: To determine the adequacy of 'in silico' biophysical models of atrial fibrillation (AF) in the design of different ablation line patterns. BACKGROUND: Permanent AF is a severe medical problem for which (surgical) ablation is a possible treatment. The ideal ablation pattern remains to be defined. METHODS: Forty-six consecutive adult patients with symptomatic permanent drug refractory AF underwent mitral surgery combined with non-transmural, (n=20) and transmural (n=26) radiofrequency Minimaze. The fraction of 'in vivo' conversions to sinus rhythm (SR) in both groups was compared with the performance of the fraction of 'in silico' conversions observed in a biophysical model of permanent AF. The simulations allowed us to study the effectiveness of incomplete and complete ablation patterns. A simulated, complete, transmural Maze III ablation pattern was applied to 118 different episodes of simulated AF set-up in the model and its effectiveness was compared with the clinical results reported by Cox. RESULTS: The fraction of conversions to SR was 92% 'in vivo' and 88% 'in silico' (p=ns) for transmural/complete ablations, 60% respectively 65% for non-transmural/incomplete Minimaze (p=ns) and 98% respectively 100% for Maze III ablations (p=ns). The fraction of conversions to SR 'in silico' correlated with the rates 'in vivo' (r2=0.973). CONCLUSIONS: The fraction of conversions to SR observed in the model closely corresponded to the conversion rate to SR post-surgery. This suggests that the model provides an additional, non-invasive tool for optimizing ablation line patterns for treating permanent AF.