Characterization of fractionated electrograms using a novel time-frequency based algorithm

Atrial fibrillation (AF) arises from complex spatiotemporal atrial activation. Current treatment for patients with AF when antiarrhythmic drugs have failed is catheter ablation which uses Radiofrequency (RF) energy to destroy heart tissues that drive AF. Therefore, AF can be terminated once the AF source is localized and eliminated by RF ablation. There is considerable interest in defining whether complex fractionated atrial electrograms (CFAE) indicate AF-perpetuation sites. This work proposes a novel time-frequency (TF) based algorithm to characterize CFAE electrograms (EGMs). The proposed technique obtains an automated classifier that is trained based on the differences evidenced between the TF structures of CFAE and non-CFAE EGMs. These characteristics are quantified using 5 TF features which are extracted using a TF matrix decomposition method performed on the EGM. The results from 5 patients with AF show that the proposed method is successful in identifying CFAE vs. non-CFAE EGMs, and might open new perspectives for a novel and reliable mapping technique to accurately characterize and understand AF mechanism.