Patient monitoring with 12-lead ECG subsets typically uses the independent frontal leads I and II and any number of the six precordial leads to reconstruct the unrecorded ECG leads. However, variations of QRS amplitudes in leads I or II may have an effect on the signal to noise ratio of the reconstructed leads. The aim of this study was to develop and evaluate a dynamic frontal lead selection method (DFLS) to improve ECG reconstruction. We compared the DFLS method for general (GEN) and patient-specific (PS) reconstruction with a lead subset I, II, V 2 , and V 5 . For GEN reconstruction, a data set of 2372 diagnostic 12-lead ECGs obtained from subjects with chest pain suggestive of acute myocardial infarction was used. For PS, a data set of 71 continuous 12-lead PCI recordings was used. Reconstruction accuracy was assessed with correlation coefficients and root mean square errors. This study showed that the DFLS method increases GEN reconstruction performance in a subgroup with low QRS voltages. PS reconstruction shows a moderate overall performance increase.