Mechanistic interpretations of sustained ventricular fibrillation: The role of mapping methodology, model and time

Abstract

Ventricular fibrillation (VF) is a malignant tachyarrhythmia underlying cardiac arrest, yet the mechanisms maintaining VF remain incompletely understood. Ethical constraints and technical limitations have largely precluded systematic investigation of sustained VF in humans, resulting in a body of evidence derived predominantly from animal models and ex vivo human preparations, often yielding apparently conflicting mechanistic interpretations. This article summarizes the experimental studies and conceptual frameworks that underpin the contemporary understanding of VF maintenance. We discuss evidence derived from frequency- and phase-based mapping approaches in short- and long-duration VF studies across species, highlighting how methodological constraints, animal models, spatial scale and VF duration critically influence mechanistic interpretation. Despite the heterogeneity across studies, overall, the data indicate that VF is not inherently chaotic but exhibits transient spatiotemporal organization that evolves dynamically over time. The underlying drivers sustaining VF throughout its progression are not uniform: multiple at times co-existing mechanisms, including re-entrant activity, self-perpetuating wavelets and rapid focal activation, have been implicated. The predominance and temporal evolution of these mechanisms are shaped by progressive global ischaemia with metabolic deterioration, pre-existing structural remodelling and the involvement as well as architecture of the Purkinje network. Collectively, these observations support the concept of duration-dependent and likely also substrate-specific electrophysiological phenotypes of VF, while also exposing a paucity of studies that have explicitly and systematically evaluated the effect of pre-existing structural and electrical substrate characteristics to examine their reciprocal interaction with VF dynamics. We conclude by outlining future directions, emphasizing the need for true 3D transmural mapping technologies and multimodal integration of structural and functional substrate features to enable mechanistic phenotyping and inform more personalized and effective strategies for VF management and prevention.

Ventricular fibrillation (VF) is characterized by rapid (>300 beats per minute), irregular electrical activation with variable electrocardiographic waveforms that prevents coordinated myocardial contraction, resulting in immediate loss of cardiac output.¹ It most commonly occurs in the context of coronary artery disease.^2,3

Resuscitation efforts are critically time-dependent: with each minute of untreated VF, the survival rate declines by approximately 6–10%, and real-world estimates place the time to first defibrillation in the pre-hospital setting between 4 and 10 min.^4–9 Yet, little is known about the mechanisms sustainingVF at these clinically relevant time points. While early (short-duration) VF has been studied during brief interruptions of coronary perfusion in humans, the arrhythmia encountered during resuscitation typically represents a metabolically more compromised and evolved state that remains poorly characterized.^10,11

Direct investigation of sustained VF in humans is fundamentally constrained by ethical considerations, and current mechanistic understanding is therefore largely derived from animal models and ex vivo human hearts. However, differences in species, substrates and experimental as well as methodological approaches have produced divergent findings that sustain an ongoing controversy regarding the dominant mechanisms responsible for VF maintenance.

At the centre of this debate lies a historical but fundamental question: Is VF sustained by a largely anarchic, disorganized process of continuously fragmenting wavelets, or does it exhibit hierarchical organization maintained by discrete electrical drivers such as re-entrant circuits or focal sources? Evidence suggests that this dichotomy may be overly simplistic.¹² Rather than representing mutually exclusive mechanisms, organized and disorganized patterns may co-exist within the same episode of VF, with their relative contributions evolving over time and being shaped by the metabolic state, underlying myocardial structure and the intrinsic properties of the Purkinje system.^12–15

In this article, we summarize the experimental evidence and proposed mechanistic concepts underlying sustained VF. We discuss the ethical, translational and technical challenges inherent to in vivo VF research and how these limitations influence mechanistic interpretation. We then examine data from short-duration VF (SDVF) in humans and long-duration VF (LDVF) in large animal or ex vivo models and the on-going controversies regarding VF maintenance. Emphasis is placed on the temporal evolution of VF, the role of the Purkinje system during prolonged ischaemia and the still only sparsely explored impact of structural remodelling on fibrillatory dynamics. Finally, we identify gaps in knowledge and outline future directions, highlighting the need for true 3D mapping and multimodal structure–function integration to enable mechanistic phenotyping ofVF and improve clinical translation.

Methodological and translational challenges in sustained ventricular fibrillation research

Mechanistic investigations of sustained VF present unique ethical, translational and technical challenges that fundamentally shape both the available evidence and its interpretation.

Ethical constraints in human ventricular fibrillation research

In vivo studies of sustained VF in humans would be unethical. VF cannot be intentionally induced and maintained for prolonged periods of time, as encountered during cardiac arrest, primarily due to the high ischaemic vulnerability of the brain.¹⁶ Consequently, available human data are largely limited to short episodes of VF studied during transient supra-coronary aortic occlusion in intraoperative settings or during invasive electrophysiological studies with VF induction.^10,17 Reported durations typically range from a few seconds to less than 1 minute with rare extensions to approximately 2 minutes.^10,11,17

While these brief episodes have provided valuable insights into early VF dynamics, they are insufficient to characterize LDVF, during which progressive metabolic derangement and ionic imbalance fundamentally alter electrophysiological behaviour.^18,19 Importantly, mechanisms such as Purkinje-mediated activity and site-specific electrophysiological behaviour secondary to differential ischaemia resistance are thought to become increasingly prominent only after several minutes of untreated VF.²⁰ Thus, human studies so far have captured only a narrow temporal window of a much more complex and dynamically evolving arrhythmia.

Alternative approaches, such as VF mapping with veno-arterial extracorporeal membrane oxygenation, have enabled longer VF recordings under ethically acceptable conditions.²¹ However, in these settings the myocardium remains perfused via retrograde coronary flow, and therefore does not reproduce the profound global ischaemia characteristic of true cardiac arrest. As a result, VF dynamics observed may differ substantially from those occurring during arrest, limiting their generalizability.

Translational considerations of ventricular fibrillation animal models

Given these restrictions, most of the mechanistic insight into sustained VF is derived from animal models. Interpretation of VF dynamics needs to take the underlying anatomical and electrophysiological properties of the selected model into account, whereas the translational value for humans will depend on their similarity to the human heart. Supplementary Table S1 summarizes electrophysiological properties of large animal models used in VF research in comparison with humans.^22–39

Most large animal VF research has been performed in dogs and pigs. Canine hearts are considered favourable for ventricular arrhythmia studies due to similarities in ventricular electrophysiological properties and Purkinje network architecture to humans.^31,32 However, substantial inter-breed variability in heart size and the presence of extensive coronary collateral circulation complicate their use as models for chronic or acute myocardial infarction.⁴⁰ In turn, porcine models are favoured for ischaemic and post-infarction studies due to similarities in heart size and coronary anatomy to humans and limited collateral circulation; yet, electrophysiological differences exist.⁴¹ Pigs exhibit more extensive transmural penetration of the Purkinje system than humans, and some repolarizing currents differ from those in humans, with altered action potential morphology, duration and restitution properties, all of which may influence propagation patterns and the stability of re-entrant activity during VF.^33–35

Additionally, most VF animal studies are conducted in young, otherwise healthy animals (see Table 1).^42–50 This fails to capture and reflect the complex and heterogeneous myocardial substrates present in patients who are at risk of VF, including post-infarct replacement scar, diffuse interstitial fibrosis, myocardial hypertrophy, the effects of metabolic diseases, such as diabetes mellitus, or the modulation by pharmacotherapy. Age- and sex-related changes are also known to influence cardiac electrophysiology and arrhythmia susceptibility but are rarely addressed.^51,52 Furthermore, the requirement for general anaesthesia in invasive studies alters autonomic tone and electrophysiological behaviour, likewise challenging the translational value of experimental VF studies for clinical events of sudden cardiac arrest.⁵³

Table 1: Overview of key terminology in ventricular fibrillation research^42–50.

Term	Definition	Surface mapping appearance	Common pitfalls and misinterpretations
Wavelet	Short-lived activation wavefront not anchored to a fixed circuit	Short linear or curved activation lines that may exist as part of a continuously changing population of wavelets (‘multiple wavelets’)	Apparent fragmentation may reflect limited spatial resolution and/or failure to account for intramural 3D propagation rather than true wave break
Wave break	Local failure of propagation due to heterogeneous refractoriness (functional), anatomical barriers and wavefront curvature or source–sink mismatch44,45	Sudden termination or splitting of an activation front	Often inferred from surface maps without accounting for intramural continuity or spatial resolution
Vortex shedding	Detachment of propagating wavefronts at the edge of an obstacle caused by wave break when local curvature exceeds a critical threshold (vortex = rotating electrical wavefront)46	Wave break with rotating/propagating wavefronts emerging at the edge of an obstacle, the same obstacle may generate new wavefronts with successive activations	A mechanism of ‘wave generation’, rather than a mechanism underlying VF maintenance
Phase Singularity	Point in a phase map where phase is indeterminate, represents sites of wave break and pivot points on which a wavefront may hinge42	Point in a phase map where phase is indeterminate, surrounded by points showing a complete phase change of 2π	PS may represent pivot points of rotors, yet, not all PS correspond to stable re-entry; may reflect transient fragmentation or colliding wavelets43
Rotor	Organized functional re-entry around an excitable but unexcited core (in 2D)47; in broader definitions the term may also include anatomically anchored, as opposed to purely functional, rotational activity	Stable or meandering rotational activity with high local frequency (stability over time is required to distinguish from transient wave break)	Presence of rotors does not automatically imply hierarchical dominance; Surface rotor may be a projection of an intramural scroll wave; Short-lived or drifting rotors may be dismissed as noise despite mechanistic relevance; Inconsistent use of the term to describe purely functional vs also anatomically anchored rotational activity
Mother rotor	Dominant (primary) high-frequency re-entrant driver generating peripheral fibrillatory conduction	Stable high-frequency rotational source with surrounding regions exhibiting wave break/block	Co-existence of multiple drivers (rotational or non-rotational) possible
Spiral wave	2D representation of the curved vortices generated by a spinning rotor48	Curved, rotating wavefronts	*Same as rotor
Scroll wave	3D extension of a spiral wave48	Paired or discordant epicardial/endocardial rotation; focal breakthroughs	Cannot be conclusively identified from surface mapping, mainly studied in computational models
Filament	Core of a scroll wave (3D), representing the longitudinal structure around which functional re-entry occurs48	Dependent on filament geometry/shape (‘I’ spanning endocardium to epicardium, ‘U’ connecting to the same surface, ‘O’ may remain intramural)	Filament geometry has so far been inferred and/or modelled mainly in computational simulations; electro-mechanical filaments have been described in experimental studies49
Driver	Localized source that plays a causal role in sustaining fibrillatory activity; its suppression or removal alters or terminates fibrillation	Site or region of functional dominance with relative temporal stability and/or higher activation rate organizing/dictating the surrounding activation patterns	Mechanistically non-specific (may be re-entrant or focal or self-perpetuating wavelets)50 Dominance inferred from surface mapping may not imply true causality
Focal source	Discrete site of impulse initiation (automatic or triggered)	Centrifugal activation pattern	May represent breakthrough from intramural scroll waves (‘pseudo-focal activation’)

PS = phase singularity; VF = ventricular fibrillation.

Technical challenges of mapping ventricular fibrillation

The ethical and translational constraints outlined above are compounded by fundamental technical challenges inherent to mapping VF itself. VF is a highly dynamic arrhythmia characterized by rapid, irregular activation, wavefront fragmentation and complex transmural interactions. Accurate mechanistic characterization would require simultaneous, global, high-resolution mapping across the entire ventricular myocardium, including the intramural space. No currently available technology fulfils these requirements.

Most in vivo studies rely on surface-based recordings from either the epicardium or the endocardium, often with only incomplete (segmental) spatial coverage (see Table 2).^{13–15,20,21,54–73} Epicardial sock arrays have been proposed to allow global, simultaneous recording; yet, they require open-chest access, are restricted to unipolar electrograms and have relatively wide inter-electrode distances for the characterization of VF wavefronts and fragmentation (for a 256-electrode sock around 8 mm).^74,75 Endocardial mapping tools, such as basket catheters, face similar limitations, with inconsistent electrode–tissue contact and low spatial density.^76,77

Table 2: Overview of existing studies on long-duration ventricular fibrillation in large animal and human models (chronological order)^{13–15,20,21,54–73}.

Publication	Year	Animal model and sample size	Mapping device	Mapping site and VF duration	Main analysis	Main finding
Worley et al.61 Am J Cardiol	1985	Dog In vivo N=26 in vivo	Five plunge needles (25 electrodes)	Transmural 20 min	VF AR - mean and SD of RR intervals for 3 s samples Biopsy at 1,5 and 10 min after onset of VF in eight dogs	Rate and regularity of CL decreased with time, first on epicardium, then deeper towards endocardium Discrete, regular, rapid deflections persisted in most subendocardial EGMs throughout the recording period
Pierpont et al.62 Am J Physiol Heart Ciro Physiol	2000	Dog In vivo N=7 HF N=6 control	Non-contact 64-electrode LV balloon (1.8x4.6)	Endo LV 5 min	Comparison of peak frequency in HF versus control	No difference in PF between HF and control Average number of simultaneous non-continuous WFs was higher in normal than in HF hearts
Huang et al.13 Am J Physiol Heart Ciro Physiol	2004	Dog In vivo N=6	504-electrode epi plaque (21x24 array), 2 mm spacing, covered 20 cm2 (20% of epi)	Epi 10 min	12 VF descriptors quantified including: PF, WF Nr, propagation velocity, fractionation, collisions, mean area activated by WF, Incidence of re-entry, focal breakthrough, block	VF was divided into five stages based on cluster analysis: Descriptor values (except block and breakthrough incidence) increased during stage I (1-11 s) and maintained high values during stage II (12-62 s) Descriptors changed quickly to values indicating greater organization, with high incidence of re-entry during stage III (63-86 s) and approached zero during stage V (311-600 s)
Moreno et al.60 Cardiovasc Res	2005	Sheep Ex vivo N=8 HF + control	Optical mapping on anterior RV and LV (4x4 cm2)	Epi 5 min	DFmax and phase singularities lasting >10 ms tracked over 5 s Four recordings perVF episode: at low pressure, 3 min after increase in pressure, after release of pressure and 5 min into global ischaemia	Low pressure: VF in HF slower (DFmax -13%) and more organized than control (PS -33%, rotors -74%) but with 20% longer life spans Acute stretch: no effect on DFmax but increased PS/rotor density similarly in both groups Ischaemia: marked decrease in DFmax, PS density and incidence of rotors in controls but not in HF
Allison et al.54 J Cardiovasc Electrophysiol	2007	Dogs N=7 Pigs N=6 In vivo	50-60 plunge needles (six electrodes), 2 mm spacing, through LV free wall	Plunge needles 10 min	SR and LDVF ATs (defined by -dv/dt) and propagation pattern Origin (earliest activation) and propagation direction determined for each wave	Predominant propagation direction in SR and LDVF: endo-to-epi in dogs, but opposite or equal directions in pigs Fastest AR: epicardial in pigs, endocardial in dogs, with increasing endo-to-epi AR gradient as LDVF progressed in dogs but not pigs
Tabereaux et al.63 Circulation	2007	Dogs Ex vivo N=6	504-electrode array (24x21,1 mm spacing)	Endo ant pap muscle 10 min	Six segments of 5 s at 0,1,3,5,7 and 10 min → 36 segments analysed AT (-dv/dt) and WF WF = if AT within 0.5 ms of each other at neighbouring site	Purkinje activations involved in the majority of WF during VF 1,018 PFs or focal WFs identified In 534 WFs, activation from working myocardium to PF In 142 WFs, activation from PF to myocardium In 342 WF, activation arising focally (PF or myocardium)
Huizar et al.57 Am J Physiol Heart Circ Physiol	2007	Pigs Ex vivo N=6	Optical mapping (ant LV, 35x35 mm, 64x64 pixels) volume-conducted ECG	Epi ant LV 10 min	Estimate variability of AP amplitude and duration, DI Median_F, WBI, RPD	Three distinct phases of local VF dynamics identified: ‘Relatively periodic’ (<1 min, high Median_F, moderate AP variability, high WBI, low RPD) ‘Highly periodic’ (1-2 min, reduced Median_F, lowAP variability, low WBI, high RPD) Aperiodic’ (3-10 min, low Median_F, high AP variability, high WBI, low RPD)
Li et al.64 Circ Res	2008	Pigs In vivo N=12 N=6 for plunge needle	504-epi array (2 mm spacing, 1.8x1.8x1.2 cm) Needle array (9x9, 6-electrode with 2 mm spacing)	Epi and transmural 10 min	AT based on -dv/dt, WF based on AT of adjacent samples 3D WF analysis → four types of WF: foci, fractionation collision, boundary (WF propagates in map) Re-entry analysis - re-entrant, multi WF, single WF	80,565 components detected, with different incidence: 0.5% re-entrant, 12.8% multiple WF, 86.6% single WF Re-entry: present early during VF but not after 3 min; complete circuit in 2D optimal plane but not on surface maps; none of the optimal planes parallel to epicardium Intramural foci: low incidence early in VF, then increase Foci and re-entry wave tips cluster at specific sites
Dosdall et al.55 Am J Physiol Heart Ciro Physiol	2008	Dog Ex vivo N=6 Lugol N=6 control	504-electrode-array endocardium + 6 plunge needles on each side of plaque	Endo (ant pap) + transmural 10 min	Endocardial ablation with Lugol solution versus control ATs based on -dv/dt Purkinje activation = identified manually	Ablation of the Purkinje system caused VF AR to slow significantly after 2 min of VF and earlier spontaneous VF termination than in controls Ablated hearts = no significant transmural AR gradient; controls = gradients after 4 min of VF
Kong et al.65 Heart Rhythm	2009	Dog Ex vivo N=6	Transmural optical (bundle of optrode at eight transmural points separated by 1.5 mm)	Transmural 10 min	ARs and APDs, Dis	AR and APD decreased during LDVF in all transmural layers → >3 min LDVF, ARs faster and Dis shorter in mid-wall and sub-endo than in epi/sub-epi No transmural differences in APD during LDVF Restitution plots showed no functional relationship between APD and DI in any layer at any stage of LDVF
Robertson et al.66 Heart Rhythm	2009	Pig In vivo N=10	504-electrode epi plaque + floating microelectrode	Epicardial + intramural 20 min	Mean APD, DI, AP Amplitude, Vmax upstroke, CL calculated from floating microelectrode recordings each minute of VF Refractory period estimated from the minimum DI	VF progression = CL, DI and Dl(min) increased versus APD, V(max) and APA decreased Increasing CL in VF is due to increased DI and not APD, which shortens. Increase in DI is due not only to increased post-repolarization refractoriness but also to poor WF propagation due to decreased APA and V max secondary to global ischaemia
Li et al.56 Am J Physiol Heart Ciro Physiol	2010	Dogs In vivo N=6	3D transmural electrode matrix (9x9x6,2 mm spacing) - area 16x16x10 mm3	Transmural LV free wall 10 min	LSVF = if none of the electrodes recorded more than one working myocardium activation and WM activation was recorded at >5% of beats	At 2.5 ± 1.8 min of VF, intramurallya more organized, synchronous activation pattern was observed - still present in 33.4% ± 31.4% during 5-10 min of VF During LSVF, PF activations always preceded working myocardium activations near the endocardium
Robichaux et al.15 J Cardiovasc Electrophysiol	2010	Dog In vivo N=6	64-electrodes basket catheter in LV, pacing from RV + 12 lead ECG	Endo RV and LV 10 min	Activation sequences of 15 successive cycles after initiation and after 1,2,3,5,7 and 10 min	<3 min LV endocardial activation complex and present throughout 78.0% ± 9.7% of each cycle = re-entry 3-7 min endocardial activation highly synchronized and present for small percentage of each cycle (18.2% ± 7.7%) = no re-entry → Synchronized activation = focally from Purkinje, LVCL>RVCL
Taylor et al.67 Am J Physiol Heart Ciro Physiol	2012	Dog Ex vivo N=28 N=9 control, N=6 in GLYB (IKATP blocker)	Epicardial optical mapping 40x40 to 65x65 mm 10 plunge needles in lateral free wall	Epi and transmural 10 min	5 s every 30 s from 0 to 10 min of LDVF in absence (control) and presence of IKATP blocker Spatial distributions of VFR, APD and DI during LDVF If asystole, pacing performed to determine excitability	VFR was the same or higher in IKATP blocker than control but did not significantly affect transmural VFR gradient IKATP blocker shortened Dis, prolonged APD and decreased percentage of inexcitable area IKATP opening during LDVF is a major factor in onset of local and global inexcitability, but has a limited role in overall deceleration of VFR and transmural VFR gradient
Kong et al.68 Am J Physiol Heart Ciro Physiol	2012	Dog Ex vivo N=6	Five intramural eight-fibre optrodes in ant LV wall towards ant. muscle	Epi, mid- and endo 5 min	Recording during SR, pacing CL 350 decreasing until loss of 1:1 capture, perfused VF and non-perfused VF Simultaneous intramural measurements of membrane voltage (Vm) and Cai2+	Vm and intracellular Cai2+ desynchronized during LDVF when not every AP was followed by a CaT (transient alternans) - desynchronization faster on epicardium CaT duration increased (+240% at 5 min), APD shortened (-17%) Vm upstrokes always preceded CaTs → spontaneous Ca2+ oscillations in myocardium not the cause of sustained VF
Venable et al.69 Am J Physiol Heart Ciro Physiol	2010	Dogs In vivo and ex vivo N=19	Plunge needles at junction RV, LV, septum In vivo Epi optical mapping ex vivo	Transmural in vivo Epi ex vivo 8 min	VFR in RV, LV, septum Optical mapping: epi distribution of VFR, APD and DI	Transmural VFR gradients developed in RV and LV with a faster VFR in endo, fastest VFR in RV-sep junction Progressive increase in VFR dispersion within LV and RV, with mosaic presence of fully unexcitable areas after 4-8 min Inverse of VFR was highly correlated with DI, but not APD → Complex VFR gradients during LDVF cannot be explained solely by distribution of Purkinje fibres
Cheng et al.70 Am J Physiol Heart Circ Physiol	2012	Pigs In vivo N=6	504-electrode arrays (ant and posterior LV) ± 40% of epicardium	Epi 10 min	Quantification of WF number, CV, fractionations, collisions, mean area activated by WF, repeatability, multiplicity index, AR, incidence of foci, block or re-entry	In pigs, only four stages of VF exist Stage I (1-22 s): values of most descriptors increased Stage II (23-39 s): values indicating greater organization Stage III (40-187 s): steady decrease of values Stage IV (188-600 s): remained relatively unchanged; epicardium still activating in stage IV
Jin Q et al.71 Heart Lung Ciro	2013	Dog In vivo N=12	2x64-electrode basket catheters in LV and RV	Endo RV and LV 7 min	DFTs measured three times in each group: SDVF (20 s), LDVF (3 min in Group A and 7 min in Group B) and the first episode of re-fibrillation after successful defibrillation for LDVF	DFT varied during different stages of VF Synchronized activation patterns on the LV endo were observed between 3 and 7 min of LDVF, but not <3 min The highly synchronized activation patterns after 3 min VF might contribute to the decreased LDVF-DFT
Li et al.14 Ciro Arrhythm Electrophysiol	2013	Dogs In vivo N=6 control N=6 pinacidil N=6 flunarizine	2x64-electrode basket catheters in LV and RV	Endo RV and LV 7 min	Quantification of organization level: Syn-idx, reg-idx Grouped into elec, synchrony versus stable versus non-stable pattern	After 3 min, VF exhibits two organized endocardial activation patterns 96% of the timePotentially arising focally in the Purkinje system - prevented with IKATP channel opener but not CCB Potentially arising from a stable re-entrant circuit near the apical left ventricular endocardium
Lin et al.72 Croat Med J	2014	Dogs In vivo N=11 (Lugol=5, control=6)	64-electrode basket catheter in LV	Endo LV 7 min	QRS duration and QT/QTc in SR, ARs in first 2 s of each minute during 7 min	All episodes of LDVF (6/6) were sustained for 7 min In Lugol’s solution group, 4/5 episodes of LDVF spontaneously terminated before 7 min (4.8 ± 1.4 min) In the control group, at 5.1 ± 1.3 min, organized focal LV endo activation pattern was observed
Huang et al.20 J Am Heart Assoc	2014	Pig and dog In vivo and ex vivo N=6, dog N=6 pig +10 dog ex vivo	504-electrode epi plaque (24x21-electrode) anterior LV + two rows of plunge (12-electrode)	Epi Transmural 10 min	In vivo AR recordings epi and transmurally Ex vivo perfused dog heart microelectrodes into Purkinje and myocardial cells	Sinus: endo-to-epi activation in dogs but not pigs Epi pacing: along endocardium and parallel activation to epi in dogs, but perpendicular to epicardium in pigs Purkinje activation in VF near endocardium in dogs but scattered transmurally in Pigs Ex vivo: >5 min Purkinje AR faster than mean working myocaridum
Jackson et al.58 JACC Clin Electrophysiol	2015	Human Ex vivo N=12	Plunge needle (25x4-electrode) and endo balloon array (112-electrode) and epi sock array (112-electrode bipoles)	Endo Epi Transmural 10 min	VF CL: median number of LATs on each needle used to define number of ‘beats’; CL = Nr of beats divided by duration of VF segment Re-entry = one full rotation with progression of local activation on at least 75% of the nine elect, spanning >85% of CL of beat	Endo-to-epi activation frequency gradient created by focal endo activations with mid-myocardial WB Focal endo activations originate most commonly from Purkinje fibres - PF shows greater resistance to ischaemia and greater capacity for continued 1:1 capture of Purkinje fibres compared with ventricular myocardium Re-entry is an uncommon mechanism in human LDVF
Panitchob et al.59 Circ Arrhythm Electrophysiol	2017	Dogs In vivo N=6	64-electrode Basket in LV 54x6-electrode plunge needles	Endo Transmural 7 min	Endocardial activation patterns: Chaotic (varying CLs and nonsynchronous activations) Regular (highly repeatable CLs) Synchronized (rapid activation spread over endocardium with diastolic periods between activations)	Chaotic pattern predominant in early VF Regular pattern emerges as VF progresses Synchronized pattern emerged occasionally in late VF →Regular and synchronized patterns of activation were driven by rapid activations on the endo surface that blocked and broke up transmurally endo-to-epi AR gradient as LDVF progresses →Failed shocks changed chaotic and regular activation to synchronized patterns in LDVF but not in SDVF
Luo et al.73 BMC Cardiovasc Disord	2017	Dogs In vivo N=6 control N=6 HF	2x64-electrode basket catheter in RV and LV	Endo 7 min	AR of VF was estimated by FFT from each electrode	Control group: LV faster than RV in first 20 s, no difference after 20 s; at 3 min post, LV fastest, ant LV slowest HF group: detectable AR LV/RV gradient within 3 min of VF - LV faster than RV, septum fastest, ant slowest
Hasegawa et al.21 HeartRhythm Case Rep	2020	Human In vivo on ECMO N=1	Endo contact mapping Perfused arrest	Endo 5 days (on ECMO)	Contact mapping → EGM, manual identification of Purkinje	Almost regular Purkinje activation preceded by ventricular activation on LV septum Purkinje ablation → mean CL prolonged → after four applications at LV apex Purkinje activity eliminated and spontaneous VF terminated approximately 1 min later

ant = anterior; AP = action potential; AFA =AP amplitude; APD = action potential duration; AR = activation rate; AT = activation time; Ca²⁺ = calcium; Ca²⁺ = intracellular calcium; CaT = calcium transient; CCB = calcium channel blocker; CL = cycle length; CV = Conduction Velocity; DP = dominant frequency; DPT = defibrillation fibrillation threshold; DI = diastolic interval; dv/dt = change of voltage over time; ECG = electrocardiogram; ECMO = extracorporeal membrane oxygenation; EGM = electrogram; endo = endocardial; epi = epicardial; EFT = fast Fourier transform; GYLB = glybenclamide; HF = heart failure; LAT = local activation time; LDVF = long-duration ventricular fibrillation; LSVF = locally synchronised VF; LV = left ventricle; Median_F = excitation frequency; Nr = number; pap = papillary; PF = peak frequency; PS = phase singularity; reg-idx = regularity index; RPD = recurrence of propagation direction; RV = right ventricle; SD = standard deviation; SDVF = short-duration ventricular fibrillation; sep = septal; SR = sinus rhythm; syn-idx = synchronicity index; VF = ventricular fibrillation; VFR = ventricular fibrillation activation rate; V_m = membrane voltage; WB = wave break; WBI = wave break incidence; WF = wavefront.

Non-contact mapping approaches, including inverse solution-based systems, have been proposed to study fibrillatory activity and enable panoramic acquisitions without direct myocardial contact.^78,79 However, their accuracy deteriorates with increasing distance from the myocardial surface, and the signal averaging required in some non-contact systems to stabilize reconstructions may obscure transient or high-frequency phenomena critical to VF dynamics.^80,81 As a result, these systems are no longer routinely in use. Similar limitations apply to electrocardiographic imaging (ECGI), which has been proposed for VF mapping but is strongly dependent on implementation choices (such as the cardiac source model and regularization method).^17,82,83 Reported performance has been suboptimal even for more organized rhythms during functional substrate mapping, raising concerns about its ability to resolve the rapid, spatially complex and temporally dynamic activation patterns that are characteristic of VF.⁸⁴

Alternative non-invasive approaches based on high-frame-xrate ultrasound circumvent the inverse problem of electrocardiographic imaging (ECGI) by inferring activation from direct tissue-based recordings of electromechanical activation patterns. These electromechanical mapping techniques have been proposed to resolve transmural activation patterns and have provided novel insight into true 3D mechanical scroll-wave dynamics during VF.^85,86 Nevertheless, the complex and frequently altered relationship between electrical and mechanical activation, particularly in diseased myocardium, may constrain their ability to accurately characterize the underlying drivers of VF in the presence of myocardial scar and fibrosis, where their performance has been shown to be less reliable even for the mapping of organized ventricular arrhythmias.⁸⁷

Lastly, optical mapping is a modality that offers unparalleled spatial and temporal resolution and has been instrumental in elucidating fundamental principles such as functional re-entry during VF.^88,89 However, optical techniques are typically limited to ex vivo preparations and frequently employ an electromechanical uncoupler, both of which have important effects on cardiac electrophysiology and are restricted to the segmental evaluation of the surface or near-surface of the myocardium.^90,91

A particularly important and unresolved limitation across all these (surface-based) approaches is the inability to directly visualize 3D activation and repolarization patterns. Distinct intramural phenomena, such as scroll waves with filaments that may intersect the surface, may manifest as focal breakthroughs or transient rotational activity on epicardial or endocardial maps. Without direct transmural information, such surface patterns are inherently ambiguous.⁶⁴ While plunge needle electrodes have been used to sample intramural activity at locally high resolution, their overall spatial coverage is limited, and the potential for mechanical and/or electrophysiological distortion by the needles themselves further complicates interpretation.^54–56

Implications for mechanistic interpretation

Collectively, these ethical, translational and technical constraints have profound implications for the interpretation of VF mapping studies. Many of the apparent discrepancies in the VF literature may not indeed represent true biological disagreement, but arise from differences in experimental context, mapping technology and the spatiotemporal scale and duration at which fibrillatory activity was interrogated. Importantly, mechanisms inferred from surface recordings may only represent a projection of more complex 3D dynamics occurring within the myocardial wall.

Recognizing these limitations is essential when integrating evidence across studies, such as those reviewed below, and restricts extrapolating findings from any single model or mapping modality. These challenges also underscore the importance of developing new experimental and analytical approaches capable of capturing VF as a dynamic, 3D process that evolves over time.

Analysis of cardiac fibrillatory activity

Interpretation of results of VF mechanisms requires a good understanding of the analytical tools and signal processing approaches employed to identify and describe electrophysiological concepts such as re-entry, wavelets and focal activity, including their limitations. Here, we outline three commonly used VF mapping approaches, frequency analysis, phase mapping and entropy-based analysis. Key mechanistic concepts and terminology used in VF research are summarized in Table 1.^42–50

Dominant frequency mapping

Traditionally, intracardiac electrograms are analysed in the time domain, plotting the amplitude of the signal against time and identifying local activation times (LATs) via peaks or maximal derivatives of the signal.⁹² In VF, electrogram amplitude and morphology fluctuate, rendering conventional LAT assignment unreliable. To address this challenge, dominant frequency (DF) analysis has been proposed.⁹³ DF analysis is applied to a given time window, dissecting the electrogram into its frequency components and creating a power spectrum based on their amplitudes. Identifying the ‘dominant’ frequency as the frequency with the highest spectral power at each recording location allows determination of the activation rate of the ventricular signals from that particular location.⁹⁴ This was an important development for evaluating fibrillatory dynamics as it facilitated the assessment of spatial organization and distribution of excitation frequencies during VF.^95,96 Figure 1 shows examples of epicardial DF maps during sustained VF. Limitations of this approach include poor temporal resolution due to the requirement of estimating DF over a time window, as well as a trade-off between the length of the time window and the resolution of the frequency spectrum. Information on the instantaneous temporal evolution of the spectral characteristics can therefore be lost and, furthermore, high-frequency drivers may be migratory, and traditional techniques may fail to account for this.⁹⁷ Some advanced frequency analysis approaches, such as the wavelet transform-based analyses, may overcome some of these limitations by enabling improved characterization of non-stationary and transient spectral features; however, they can be computationally more intensive.⁹⁸

Figure 1: Dominant frequency maps.

DF mapping allows to evaluate the spatial distribution of excitation frequency during VF across the mapped surface without relying on conventional LAT assignment, and to localize high-frequency sites that may drive VF. The figure shows examples of epicardial DF maps derived from an epicardial sock array in an in vivo porcine model and visualized as 2D polar plots (centre = apex; upper hemisphere = anterior surface; lower hemisphere = inferior surface). Top: Early VF is characterized by widespread rapid activation rates (red). As VF progresses, spatial heterogeneity in DF increases and localized sites of higher frequency (t3–t4) emerge, suggesting a more hierarchical organization. Bottom: Later stages of VF are marked by an overall decline in activation rate (green/blue), yet local sites of high-frequency drivers persist (t5–t7). The myocardium becomes increasingly unexcitable, and in the very later stages of VF, the residual ‘high’-frequency drivers are progressively slowing (t8). Figure 1 plots were created in MATLAB (MathWorks^®, Natick, MA USA) and arranged in and exported from Microsoft PowerPoint (Microsoft, Redmond, WA, USA)

DF = dominant frequency; LAT = local activation time; VF = ventricular fibrillation.

Phase mapping

Phase represents a fundamental characteristic of a signal and may be used to describe repetitive or oscillatory processes, where the state of the system is represented as a phase angle (θ).⁴² In the context of VF mapping, each point on the myocardial surface can be assigned a phase angle, and the progression of θ is then tracked over time.⁴³ One of the main advantages of phase mapping is that it does not rely on the annotation of activation times from complex and changing low-amplitude electrograms to construct activation maps, but, instead, analyses the distribution of phase to assess underlying activation patterns.⁴²

To track wavelets and detect and characterize rotational activity, the phase of all electrograms that are acquired over the epicardium and/or endocardium is analysed at every instant of time with high temporal resolution. Phase singularities are points surrounded by a complete cycle of phase and so are connected to a broken wavefront and can be the pivot of rotational activity. Detected phase singularities can then be tracked in the spatiotemporal phase maps to define their lifetime and mobility (Figure 2).⁴³ The development of phase mapping was crucial for describing rotational activity as one of the underlying mechanisms of VF.⁸⁸

Figure 2: Phase maps and corresponding unipolar voltage maps.

The development of phase-mapping techniques to study fibrillatory dynamics has been fundamental for VF research, notably for the detection and characterization of rotational drivers. Top: Examples of sequential screenshots (t1–t5) of epicardial phase maps that allow to track rotational activation wavefronts pivoting around phase singularities progressing through a complete phase change (2π). Rotors may be stationary or migratory. Bottom: Corresponding unipolar voltage maps for each phase map, with superimposed phase singularities (marked as green dotes) and wavefronts (plotted as blue lines). White dots represent electrode locations. Individual plots were made in MATLAB (MathWorks^®, Natick, MA USA) and Microsoft PowerPoint (Microsoft, Redmond, WA, USA) for arranging the plots together and exporting

Uni = unipolar; VF = ventricular fibrillation.

Entropy-based analysis

In addition to frequency- and phase-based approaches, measures from information theory have been applied to quantify the regularity and complexity of fibrillatory activity.⁹⁹ Entropy is defined as a measure of ‘disorder or uncertainty in a system’, and entropy-based metrics of electrophysiological signals have been used to quantify the (ir) regularity or complexity of VF waveforms.¹⁰⁰ Unlike phase mapping, entropy-based measures do not explicitly identify activation patterns per se, but rather provide a local or global assessment of the underlying ‘state’ of VF, thereby distinguishing more organized from more disorganized fibrillatory activity. A main research interest has been to apply entropy metrics to surface-electrocardiography (ECG) VF waveforms to predict defibrillation outcomes in cardiac arrest, rather than to directly identify mechanistic VF drivers.⁹⁹

Multiple forms of entropy analysis have been proposed, each capturing slightly different aspects of signal regularity and structure. For example, 'Shannon entropy' is based on amplitude distribution and quantifies its ‘unpredictability’, reflecting signal heterogeneity (Figure 3).¹⁷ In turn, for example, ‘fuzzy entropy' assesses the similarity of temporal patterns over time, capturing the temporal predictability of the signal.¹⁰¹

Figure 3: Entropy estimation.

The top-left panel shows a segment of the raw VF electrogram. Subsequent subpanels on the left illustrate the application of sliding temporal windows over the signal. Each window represents a short segment of the VF waveform over which signal complexity is quantified; sliding windows are overlapping to provide a continuous estimate of signal organization over time. In the middle panel, two examples of amplitude histograms from the raw VF signal are shown using 16 bins. Shannon entropy measures the uncertainty or irregularity of a signal by quantifying the distribution of amplitude values within each window: the maximum possible entropy occurs if all bins are equally likely (maximum uncertainty). The numeric output of entropy is relative to the number of bins used – in this example, both windows show relatively high signal complexity. On the right, the raw VF signal is plotted with Shannon entropy shown as shaded colours overlying the respective time window, illustrating how signal organization varies over time (only consecutive windows are coloured to facilitate interpretation of colours without overlap). Graphs were made in MATLAB (MathWorks^®, Natick, MA, USA) and compiled together and exported from Microsoft Powerpoint (Microsoft, Redmond, WA, USA)

VF = ventricular fibrillation.

A common feature of entropy metrics is that they are calculated on electrogram time series within defined temporal windows. In general, higher entropy values indicate increased signal complexity and reduced temporal organization, while lower values reflect more repetitive or organized activity. In turn, limitations of entropy-based analysis include sensitivity to parameter choice, with a notable absence of guidelines on their selection, dependence on signal length and noise characteristics and limited spatial specificity.⁹⁹

Frequency-, phase- and entropy-based mapping approaches should be considered complementary rather than stand-alone tools for VF analysis, with each extracting distinct properties from fibrillatory waveforms.

Mechanisms sustaining ventricular fibrillation: concepts and evidence

Ventricular fibrillation as a dynamic evolving process

A consistent observation across experimental studies was that sustained VF is not a mechanistically uniform arrhythmia but evolves dynamically over time, as metabolic conditions deteriorate during global myocardial ischaemia, which alters electrophysiological properties. This temporal evolution was recognized nearly a century ago.¹⁰² In seminal canine experiments employing high-speed cinematography and electrocardiography, Carl Wiggers described four sequential stages of VF (I=‘undulatory’, II=‘convulsive’, III=‘tremulous’, IV=‘atonic’), characterized by progressive slowing of activation rate and declining mechanical activity as ischaemia progresses. Subsequent experimental work built on this concept, proposing three-, four- or five-stage VF models depending on species, mapping methodology and analytical metrics.^13,57,103 For example, Huang et al. described, also in dogs, a rapid increase in VF metrics (such as number of wavefronts, fractionations and incidence of re-entry and focal activation) within the first seconds of VF (1–11 s), followed by rapid dynamic fluctuations during stage II (12–62 s) and greater organization in stage III (63–86 s), which thereafter decreased steadily in stage IV (87–310 s) prior to approaching zero during stage V (311–600 s).¹³

In contemporary literature, VF is often more pragmatically divided into SDVF, typically lasting <1–3 minutes, and LDVF, extending beyond this period.^12,58 While somewhat arbitrary, this distinction reflects a fundamental feature of VF: the arrhythmia actively and rapidly alters its own substrate secondary to the immediate loss of cardiac output at VF onset. SDVF occurs under relatively preserved metabolic conditions, whereas LDVF unfolds in the context of progressive global ischaemia with increasingly impaired excitability and heterogeneity in refractoriness.¹⁸ These changes significantly influence VF dynamics and mechanisms maintaining it.

Recognizing VF as a time-dependent process is critical, as mechanistic conclusions drawn from SDVF cannot be directly extrapolated to later stages encountered during clinical cardiac arrest.

Competing frameworks: anarchy versus hierarchy

Wave break, caused by a localized conduction block at a specific site along the wavefront while other portions continue to propagate, is an established hallmark of cardiac fibrillation.¹⁰⁴ Possible mechanisms that may result in wave break are anatomical barriers, heterogeneity in refractoriness in the tissue, critical wavefront curvature and source–sink mismatch.⁴⁸ Yet, the exact mechanisms that sustain and perpetuate the on-going wave breaks during VF, especially in its later stages, remain an area of on-going debate. Figure 4 illustrates a schematic of different proposed mechanisms underlying sustained VF.

Figure 4: Overview of proposed mechanisms sustaining ventricular fibrillation.

Multiple mechanisms may co-exist and/or interact within a single episode of VF. The dominant mechanism can vary with VF duration, metabolic state and pre-existing structural alterations, and may shift dynamically over time. Figure 4 was created in Microsoft PowerPoint (Microsoft, Redmond, WA, USA)

VF = ventricular fibrillation.

Historically, the debate regarding VF maintenance has centred on two seemingly opposing conceptual frameworks. The ‘anarchical’ multiple wavelet hypothesis proposes that cardiac fibrillation is sustained by numerous, continuously fragmenting self-perpetuating wavelets that propagate through the myocardium along constantly changing pathways, with fibrillation persisting as long as sufficient myocardial mass allows the continued generation of daughter wavelets.¹⁰⁵ Experimental mapping studies in animal models that have recorded highly complex activation patterns with frequent wavefront fragmentation in the absence of a primary rotational driver supported this view.^106,107

In contrast, the ‘hierarchical’ or mother rotor hypothesis posits that VF is organized around one or more discrete high-frequency sources, typically functional re-entrant circuits, that act as drivers.¹⁰⁸ From these drivers, wavefronts propagate into surrounding myocardium with longer refractory periods, resulting in intermittent conduction block and spatially complex activation patterns, often described as ‘fibrillatory conduction’. Optical and electrical mapping studies demonstrating relatively stable, high-frequency rotational activity with surrounding disorganization have provided supporting evidence for this framework.^88,96

Based on these apparently contradictory experimental findings, some studies suggested that indeed there may be two different types of VF: type 1 (fast) VF, associated with steep action-potential-duration restitution, flat conduction velocity restitution and multiple wavelets, versus type 2 (slow) VF, associated with flat action-potential-duration restitution, decreased excitability and spatiotemporal periodicity in activation maps.^109,110 However, on the other hand, they may also represent different and/or interacting manifestations of the same underlying dynamics observed at different spatial scales, anatomical locations and time points, and/or indeed may co-exist.¹⁰ Rapid re-entrant sources can generate wavefronts that fragment as they encounter heterogeneities, giving rise to multiple wavelets that may self-perpetuate. Conversely, wavelet interactions can themselves create new sites of re-entry that may stabilize and result in new frequency-dependent fragmentation into multiple wavelets.⁵⁰ Thus, VF may not be sustained by either mechanism alone, but rather by their continuous interaction.

The more contemporary VF debate has increasingly incorporated computational modelling and in silico studies alongside experimental investigation. While the present article focuses primarily on experimental evidence, computational studies have both helped to interpret experimental findings as well as provided important complementary insights by exploring possible electrophysiological dynamics that are difficult to capture experimentally.¹¹¹

Insights from computational modelling

With respect to the historical ‘anarchy versus hierarchy’ debate, computational studies have demonstrated that both multiple wavelet behaviour and organized high-frequency re-entrant drivers may emerge from the same underlying substrate depending on restitution slope, conduction properties and tissue heterogeneity.^112–115 These findings support the concept discussed above that hierarchical and self-perpetuating wavelet frameworks are not mutually exclusive mechanisms, but rather represent different manifestations of non-linear wave dynamics observed under variable conditions and substrates.

A particularly important contribution of modelling lies in resolving the inherent ambiguity of surface-based mapping. 3D simulations have shown that intramural scroll waves with filaments extending through the ventricular wall may drive VF, and may manifest on the epicardial or endocardial surface as focal breakthroughs, spiral waves or apparently disorganized activation patterns.^116,117 Surface activation maps do not necessarily provide a complete representation of the underlying mechanism, as they may reflect a projection of organized intramural dynamics, which currently escape experimental techniques but can be explored in computational models.

Computational approaches have also provided insight into the time-dependent electrophysiological changes secondary to acute myocardial ischaemia, including reduced excitability, conduction slowing, altered refractoriness and gap junction uncoupling, all of which may modulate VF dynamics.^18,118 Such frameworks may help inform more realistic VF simulations accounting for the time-dependent evolution of the arrhythmogenic substrate under deteriorating metabolic conditions. Lastly, incorporation of an anatomically detailed His–Purkinje system has enabled the evaluation of the potential contribution of the specialized conduction system to the initiation and maintenance of VF.¹¹⁹ The combination of realistic anatomical models, dynamic metabolic perturbations and ischaemia responses may help explain transitions between re-entry-dominated and focal or Purkinje-mediated activity observed experimentally during LDVF.

At the same time, it is important to emphasize that models ‘only’ approximate reality and rely on assumptions and simplifications regarding tissue structure and the mechanisms underlying the action potential. Most existing modelling frameworks do not capture the biophysical details of individual myocytes, and neither do they include the dynamic metabolic and autonomic alterations that characterize prolonged VF. Modelling studies have to balance the level of detail against the computational cost of simulations, with model complexity chosen according to the type of study that is being undertaken. Consequently, in silico findings should be interpreted within the context of their underlying parameterization and complexity. Nevertheless, the detailed integration of experimental observations with computational modelling offers a powerful tool to interrogate mechanisms that, for now, remain inaccessible to direct measurements in vivo.

Evidence from human short-duration ventricular fibrillation

Intra-operative epicardial mapping studies during brief episodes of VF have demonstrated that early human VF often exhibits a higher degree of organization than that observed in animal models of comparable heart size.^10,11 These studies reported activation patterns consistent with both re-entrant activity and wavelet propagation, suggesting co-existence rather than exclusivity of mechanisms.

High-resolution optical mapping studies in explanted Langendorff human hearts have further supported this view.¹²⁰ Analyses of brief VF segments revealed relatively slow conduction velocities (CV), around 0.25 ± 0.04 m/s during VF compared with CV_max (maximum conduction velocity) of 0.87 m/s during pacing, and the presence of large, coherent wavefronts spanning wide regions of myocardium. Compared with canine or porcine hearts, human VF displayed longer activation cycle lengths and fewer simultaneous wavefronts. This has been attributed to longer minimum action potential durations in humans, which increase wavelength and may limit activation rate and the number of sustainable re-entrant circuits or wavelets at any given time.³⁶

SDVF studies combining epicardial and endocardial mapping in patients with cardiomyopathy have provided additional insights.¹²¹ Short episodes of VF were often characterized by wavefronts emanating from a limited number of sites, frequently localized to endocardial regions of low voltage, a surrogate for scarred tissue. This was supported by histopathological correlation in explanted human hearts with dilated cardiomyopathy, demonstrating greater fibrosis at sites of conduction block and re-entry during brief episodes of VF, suggesting a role for structural heterogeneity in anchoring or stabilizing re-entrant activity.¹²² Lastly, simultaneous endocardial–epicardial recordings have also revealed rotational activity of similar cycle length and chirality on both surfaces, consistent with transmural re-entry in the form of scroll waves, with histopathology also showing greater fibrosis at sites of re-entry than in regions without.¹²³

Notably, across these human SDVF studies, repetitive focal firing from the Purkinje system was not observed.

Long-duration ventricular fibrillation: insights from animal models and ex vivo human hearts

Mechanistic investigation of in vivo LDVF has only been possible in large-animal models, but not in humans. These studies revealed substantial differences in VF drivers when compared with early VF, particularly with respect to activation rate gradients, endocardial–epicardial dissociation and the emergence of focal activity. A compilation of published studies investigating LDVF and their main findings in large animals or ex vivo human hearts is presented in Table 2 in chronological order.^{13–15,20,21,54–73}

A recurring observation during LDVF is the development of a pronounced endocardial-to-epicardial activation rate gradient, with faster rates recorded near the endocardium. This phenomenon has been most consistently observed in canine and ex vivo human hearts and has been strongly linked to the Purkinje system.^15,58,59 In turn, chemical ablation of the Purkinje system abolishes the development of an endocardial–epicardial gradient.⁵⁵ Purkinje fibres exhibit relative resistance to ischaemia, attributed to higher glycogen content and partial perfusion from the ventricular cavity, which may enable them to remain excitable when the working myocardium has already become severely compromised.^124,125

Most of these studies have recorded repetitive, rapid Purkinje activation preceding myocardial activation during LDVF, consistent with a focal driver role. Proposed mechanisms include abnormal automaticity or triggered activity mediated by calcium-dependent afterdepolarizations under metabolic stress.¹²⁶ Importantly, Purkinje-driven activity does not exclude concurrent re-entry. Rather, focal excitation may interact with residual re-entrant circuits or generate wavefronts that fragment upon encountering heterogeneous tissue, contributing to on-going fibrillatory conduction. Whether, or not, in the very late stages, when re-entrant and self-perpetuating wavelet mechanisms may no longer be viable due to widespread conduction failure, they may represent the sole driver of on-going electrical activity remains unclear.

Ventricular fibrillation dynamics in structural heart disease

Despite numerous experiments (Table 2), LDVF studies integrating electrophysiological observations with underlying myocardial structure are limited. Although myocardial scar and associated electrophysiological alterations are well-established determinants of organized macro-re-entrant ventricular tachycardia, their specific contribution in sustaining and modulating VF, particularly during later stages of VF, remains poorly defined.^127–129

Most experimental LDVF studies have been conducted in structurally normal hearts, with relatively few using models of pacing-induced heart failure.⁶⁰ The small number of investigations that have incorporated structural assessment alongside electrophysiological mapping typically recorded only SDVF segments and relied on limited histological sampling in ex vivo conditions or indirect inference using voltage mapping (see above).^121–123 In turn, data from small-animal perfused ex vivo models suggest that fibrosis patterns and gap junction coupling influence activation patterns during prolonged fibrillation, but marked interspecies differences between rodents and humans and the methodological approach (stable perfused condition) prevent direct translation to human cardiac arrest.¹³⁰

Future directions

Collectively, the evidence reviewed supports a view of VF as a dynamic arrhythmia with multiple interacting and/or co-existing processes whose relative contributions evolve over time, resulting in intermittent spatiotemporal organization and disorganization. Apparent discrepancies among existing mechanistic studies may reflect differences in the investigated VF duration, experimental model, spatial resolution and recording site, as well as signal processing techniques. The historical dichotomy between multiple wavelet and hierarchical driver hypotheses is obsolete and fails to capture the complexity of activation patterns observed during VF.

A central unresolved question is the role of underlying myocardial structure as an active determinant of VF dynamics. Structural remodelling, including scar architecture, heterogeneity, altered fibre orientation, hypertrophy, etc., modulates conduction–repolarization properties, source–sink balance and preferential anchoring of re-entrant circuits, thereby may mechanistically shape VF maintenance and influence the response to interventions to terminate it.¹²⁷ This gap in knowledge is particularly relevant given the diversity of substrates predisposing to and underlying VF in patients, including chronic infarction, inherited and acquired cardiomyopathies, acute ischaemic events or primary electrical diseases. Sustained VF dynamics are unlikely to ever be fully understood without explicit consideration of the underlying substrate.

The time-dependent and likely substrate-specific effects support the need for mechanistic phenotyping of VF. Rather than treating VF as a uniform entity, this framework would aim to identify dominant sustaining processes and their temporal evolution within defined clinical disease models. Such mechanistic phenotypes may differ in their susceptibility or respective response to defibrillation, pharmacological and/or localized ablative interventions, and their propensity for recurrence. For example, VF sustained predominantly by Purkinje-mediated focal activity during prolonged ischaemia may respond differently to anti-arrhythmic drugs and recurrence after defibrillation than VF driven by stable rotors anchored to the border zone of fibrotic myocardium. While such concepts remain largely hypothetical at present, they are supported by experimental observations and established principles of cardiac structure–function–electrophysiology relationships.

Accurate mechanistic phenotyping will require advances in technologies offering true 3D transmural mapping in vivo to distinguish surface manifestations from intramural dynamics (Figure 5). Equally, multimodal investigations integrating electrical, structural and potentially also metabolic and autonomic data are necessary to more comprehensively capture the full complexity of VF dynamics. Third, future experimental models must better reflect the diversity of human disease associated with VF, including clinically relevant patterns of remodelling as well as integration of modulating factors such as age, gender, comorbidities and pharmacotherapy. Experimental work should inform and be complemented by computational frameworks that may be better suited to process such high-dimensional, complex datasets and may provide a valuable tool to explore therapeutic interventions in silico before translation to experimental and clinical studies, including evaluation of anti-arrhythmic drug effects, modulation of gap junction coupling or targeting of Purkinje-mediated activation.

Figure 5: Surface projection versus ‘true’ 3D ventricular fibrillation dynamics.

A major shortcoming of contemporary mapping technology is the absence of true 3D mapping. Surface projection of intramural phenomena, such as scroll waves, may be ambiguous and may prevent accurate interpretation of VF dynamics when evaluated by surface mapping alone. Left: Schematic example of an I-shaped filaments spanning across the myocardial wall and manifesting as 2D rotational surfaces on both sides. Right: Schematic examples of two U-shaped filaments which link rotors with clockwise and counterclockwise rotation on the same surface are shown. On the opposing surface, the activation wavefront breaks out with a centrifugal activation pattern mimicking a focal source. Alternative filament geometries have been described, such as L-shaped (connected to one surface, not shown) or O-shaped (may be entirely intramural, not shown). Figure was created in Microsoft Powerpoint (Microsoft, Redmond, WA, USA)

PS = phase singularity; VF = ventricular fibrillation.

Therapeutic implications of ventricular fibrillation phenotyping

In the acute setting of cardiac arrest, an improved substrate-specific understanding of VF mechanisms may help explain time- and potentially driver-dependent variability in defibrillation success and provide a framework for future refinements in defibrillation strategies, including the timing of shock delivery and waveform design tailored to the VF stage and dominant electrophysiological drivers. Recognition of distinct VF dynamics may further inform adjunctive treatment approaches, such as the timing and selection of anti-arrhythmic drug administration and/or the development of mechanism-directed pharmacological interventions aimed at promoting a transition from VF to more organized activation restoring cardiac output.

Beyond the acute phase, mechanistic phenotyping of VF and identification of substrate-specific drivers may contribute to improved management of survivors of cardiac arrest and inform more personalized preventive strategies. VF sustained by rotor-dominated activity anchored to structural alterations, for example, the scar border zone, or by focal Purkinje firing may represent phenotypes amenable to substrate-guided catheter ablation to prevent the maintenance of VF. Furthermore, linking VF mechanisms to underlying myocardial substrate properties could also facilitate the identification of high-risk phenotypes susceptible to sustained VF and support risk stratification, potentially also guiding primary preventive interventions.

While these therapeutic strategies currently remain largely conceptual, integration of advanced electrophysiological mapping, high-resolution imaging and computational modelling may provide a pathway towards mechanism-informed, and ultimately more personalized, approaches to VF management and prevention.

Conclusions

Mechanistic understanding of sustained VF has advanced substantially over recent decades, yet remains incomplete due to fundamental ethical, translational and technical constraints in studying VF. Existing evidence across species and experimental approaches converges on several key principles: VF is not chaotic but displays intermittent degrees of organization; multiple sustaining mechanisms may co-exist within a single episode; and these mechanisms evolve dynamically as metabolic conditions deteriorate. Myocardial structure likely exerts a critical influence on fibrillatory dynamics, yet studies evaluating substrate-specific mechanisms are sparse.

Recognizing VF as a dynamic, time- and substrate-dependent process rather than a single mechanistic arrhythmia reconciles much of the apparent controversy in the literature and provides a conceptual framework for future investigation. Continued progress will depend on the development of true 3D mapping technologies, the integration of structural and functional data, and a shift towards substrate-specific mechanistic phenotyping. Advances in these areas may allow improvements in the prevention and treatment of VF and ultimately enhance outcomes in patients with cardiac arrest.

Funding Statement

Support: Johanna B Tonko is supported by an IPPRF (Imperial Post-doc, Post-CCT Fellowship), Vineesh Kappadan and Fu Siong Ng receive support from British Heart Foundation RG/F/22/110078. Ahmed El-Medany is supported by a British Heart Foundation (BHF) clinical research training fellowship (FS/CRTF/25/24752). Hesham Aggour is supported by BHF CRTF (FS/CRTF/25/24834). Aidan Birdi is supported by a BHF MBPhD studentship (FS/MBPhD/25/28044). Jean-Baptiste Guichard acknowledges support from 'Contractes d'Investigació Avançada Fundació BBVA – Hospital Clínic Barcelona Joan Rodés – Josep Baselga 2022' (HCB_BIO_001/2) and the Research Grant Daniel Bravo 2024 (Foundation Daniel Bravo), Joseph Barker is supported by a BHF grant (FS/CRTF/24/2462).No funding was received in the publication of this article.