OBJECTIVE: The surgical Maze III procedure remains the gold standard in treating atrial fibrillation (AF); however due to clinical difficulties and higher risks, less invasive ablation alternatives are clinically investigated. The present study aims to define more efficient ablation patterns of the modified maze procedure using a biophysical model of human atria with chronic AF. METHODS: A three-dimensional model of human atria was developed using both MRI-imaging and a one-layer cellular model reproducing experimentally observed atrial cellular properties. Sustained AF could be induced by a burst-pacing protocol. Ablation lines were implemented in rendering the cardiac cells non-conductive, mimicking transmural lines. Lines were progressively implemented respectively around pulmonary veins (PV), left atrial appendage (LAA), left atrial isthmus (LAI), cavo-tricuspid isthmus (CTI), and intercaval lines (SIVC) in the computer model, defining the following patterns: P1=PV, P2=P1+LAA, P3=P2+LAI, P4=P3+CTI, P5=P3+SIVC, P6=P5+CTI. Forty simulations were done for each pattern and proportion of sinus rhythm (SR) conversion and time-to-AF termination (TAFT) were assessed. RESULTS: The most efficient patterns are P5, P6, and Maze III with 100% success. The main difference is expressed in decreasing mean TAFT with a correlation coefficient R=-0.8. There is an inflexion point for 100% success rate at a 7.5s TAFT, meaning that no additional line is mandatory beyond pattern P5. CONCLUSIONS: Our biophysical model suggests that Maze III could be simplified in his right atrial pattern to a single line joining both vena cavae. This has to be confirmed in clinical settings.