Abstract : Radiofrequency catheter ablation (CA) is increas- ingly employed to treat atrial fibrillation (AF), yet se- lection of patients who would actually benefit from this therapy still remains an open issue. The present work introduces some non-invasive quantitative parameters to discriminate between successful and failing CA pro- cedures by exploiting the spatial diversity of the 12-lead surface ECG. They are based on the normalized mean square error (NMSE) between consecutive atrial ac- tivity ECG signal segments and their rank-n approx- imations determined by principal component analy- sis. As opposed to the single-lead approach of previous works, we consider NMSE values computed on more than one lead. One such multilead-based parameter is able to distinguish between successful and failing abla- tions before performing the procedure with AUC=0.86, p value=7*10−5. This study demonstrates that the pro- posed multilead parameters can effectively predict CA outcome and potentially contribute to more accurate patient selection strategies for this AF therapy.