Introduction Ventricular fibrillation (VF) is the main cause of sudden cardiac death. Despite its clinical significance, there is still a lack of quantitative tools to evaluate the degree of myocardial electrical organization during VF. This study puts forward novels markers of VF complexity and stability on body surface potential maps (BSPMs), uantifying the ability of signal features determined through principal component analysis (PCA) to be retrieved along the recording. Methods Biventricular potential mapping was performed through a 252-electrode vest in 27 VF patients (25 male, 49 ± 22 years). BSPMs were divided in 0:5-s segments, each projected on a 3D subspace determined through PCA in the preceding segment. At each time instant, the multilead error E between the input signal and its PCA approximation was expressed in terms of normalized amplitude norm |E| and angle cos(E). Additionally, the nondipolar component index (NDI) was determined as the fraction of energy retained by the first 3 PCA eigenvalues in sliding windows of the same duration. In 24 VF recordings (duration 19.8±6.5 s), average values of the aforementioned indices were computed in 4-second windows at the beginning and at the end of the output parameter series. Temporal changes in VF complexity were verified through an unpaired Student’s t-test. The same analysis was performed in 5 control recordings in sinus rhythm (SR). Results A significant increase in spatiotemporal complexity was observed during VF, quantified by higher NDI and |E| and lower cos(E) values at the end of the episode (p < 0:0001). By contrast, no significant changes between the beginning and the end of the recording were observed during SR (p > 0:05). Conclusions This study demonstrates the ability of PCA to capture and quantify changes in signal complexity during VF in a noninvasive framework.