Non-invasive assessment of spatiotemporal organization of ventricular fibrillation through principal component analysis

Introduction Ventricular fibrillation (VF) is the main cause of sudden cardiac death, but we lack tools to predict the evolution of its complexity. This study proposes novel VF complexity markers obtained by principal component analysis (PCA) of body surface potential maps (BSPMs). Methods BSPMs were divided in 0.5-s segments, each projected on the 3D PCA subspace determined in the previous frame. Reconstruction error ϵ was expressed in terms of norm d
_ϵ
and angle cosine cos(α
_ϵ
), and the nondipolar component index (NDI) quantified the energy of the first 3 PCA eigenvalues. These markers quantified changes in complexity between the beginning and the end of 24 VF episodes and 5 control sinus rhythm (SR) recordings. They were also tested on 6 BSPMs from a torso-tank model during and after ex-vivo pig heart perfusion. Differences between in-silico BSPM right and left leads were then verified in 4 human VF simulations, with a reentrant source in the right ventricle. Results Higher NDI and d
_ϵ
and lower cos(α
_ϵ
) denoted higher complexity at the end of VF (p <; 0.0001). No changes occurred in SR (p > 0.05). The indices also underlined higher disorganization in ex-vivo non-perfused VF (p <; 0.0001). A similar trend was observed in the right in-silico BSPM leads (p <; 0.05). Conclusions PCA can non-invasively quantify changes in VF complexity.