Research

For a list of publications related to DGM, please check the publications section at the bottom of this page.

DGM2,5 stands as a versatile tool capable of analyzing various arrhythmias, including Atrial Tachycardia (AT), Ventricular Tachycardia (VT), Atrial Fibrillation (AF), Ventricular Fibrillation (VF), and Torsade de Pointes (TdP). Operating on a unique principle, DGM transforms arrhythmia datasets into directed networks. It efficiently processes files containing spatial coordinates and simultaneously measured intra-cardiac signals or local activation times (LATs) from electrodes.

No regularity in spatial dataset locations is required, making DGM independent of the measuring system. It accepts input from computational, experimental (such as needle or socket data), or clinical datasets (like grid electrode or basket catheter data). The spatial locations of electrodes become the network nodes. DGM then generates a directed network representing the arrhythmia, enabling in-depth analysis to study and potentially unveil the underlying mechanisms.

DGM is recognized for its capability in analyzing reentry loops within arrhythmias, including functional reentry (rotors) and anatomical reentry (micro and macro). To reveal these reentry loops, DGM searches for closed cycles within the directed graph. Additionally, DGM can also identify focal sources, in essence recognized as nodes with only outgoing arrows. We are currently in the process of integrating other network measures to further unveil additional features of arrhythmias.

Below you can find an overview of our past and current studies about AT, AF, VT, VF, TdP and some more general studies comparing DGM with phase mapping.

Atrial tachycardia · AT

Reentry is an underlying mechanism in many cases of atrial tachycardia (AT). While the prevalent belief is that AT is mostly driven by a single reentry loop, we hypothesize that this is usually not the case. The key factor behind this lies in the index theorem, prompting a deeper exploration into the atrial topology for a more nuanced understanding.

Anatomy and Holes: topology of the left and right atrium

Considering the atria as closed surfaces with holes or boundaries, we can, topologically speaking, conceptualize (deform) the LA and RA as spheres with holes. The anatomy of the LA and RA reveals at least three natural openings or holes. For the LA, these include the mitral valve (MV), the left pulmonary veins (LPV), and the right pulmonary veins (RPV). Correspondingly, the RA features openings such as the tricuspid valve (TV), the inferior vena cava (IVC), and the superior vena cava (SVC). Additionally, patients may have non-conductive scar tissue, which, if not connected to natural openings, behaves as additional holes, altering the atrial topology. Consequently, we can uniquely identify a patient's atrial topology by counting the number of holes in the atria.

It is crucial to note that ablation lines or natural lines of block connecting two holes effectively reduce them to a single topological hole.

Transformation of left atrium to sphere
Index theorem

The index theorem states that the sum of topological charges of all loops on a closed surface (like a sphere) with a finite number of holes should be zero (Davidsen et al, 2004) . This has profound implications for AT, as the index theorem suggests that reentrant loops around holes will always come in pairs and the number of relevant reentrant loops, that need to be identified for correct treatment, will dependent on the topology of the respective chamber.

In a 2-hole model, paired rotation essentially mirrors the clockwise and counterclockwise projection of a single reentrant circuit. In the 3-hole model, paired rotation indicates two clinically relevant loops around two holes (index +1, -1) with fused activation (index 0) encircling the third hole. This principle extends similarly to models with more than three holes in the atria.

True and suppressed loops

Although one reentry loop will always be obvious as it makes a full rotation (a true loop), the second reentry loop might be suppressed due to a collision with the true loop, and therefore will not be able to make a clear full rotation. However, calculating the index around a hole with a suppressed loop will result a ± 1. Therefore, this reentry is equally relevant as a true loop. We have performed over 1000 computer simulations in 2-hole, 3-hole and 4-hole models which all confirmed our hypothesis.

Ablation

The direct termination of tachycardia can only be achieved by ablating the two holes with + and – indices. Failure to ablate the suppressed loop will result in the emergence of a slower AT. This explains why slower ATs often emerge after the first ablation line.

Ablation strategy
Unique classification of AT

From the index theorem follows that we can uniquely categorize an AT by identifying the number of holes and calculating the index of the hole.

DGM-TOP

We expanded the capabilities of DGM from identifying only true loops to the enhanced version, DGM-TOP. Unlike its predecessor, DGM-TOP can identify both true and suppressed loops. To operate, DGM-TOP requires input data from CARTO, RHYTHMIA, or openCARP simulation files, complete with cut-outs of the holes. DGM-TOP then accurately identifies the index of each hole and recommends the optimal ablation line for effective intervention.

True and Suppressed reentry
Benefits for the patient

We showed for the first time that suppressed loops play a pivotal role as not ablating them merely leads to slower AT. Due to a full understanding of the mechanism of AT from the first map benefits include operator-independent analysis, a reduction in the need for redo maps, and consequently, shorter procedure times.

Clinical proof

Our work on AT is performed in collaboration with Prof. Mattias Duytschaever and Prof. Sebastian Knecht from the hospital AZ Sint-Jan, Bruges, Belgium, who work with the CARTO system (Biosense Webster) and Dr. Armin Luik and Eng. Annika Haas from the Städtisches Klinikum Karlsruhe, who mainly work with RHYTHMIA (Boston Scientific). We are in the process of analyzing over 170 AT cases.

Further reading

In previous work, we analyzed the true loops of complex ATs 2 3. Also, an editorial comment (John M. Miller and Tanyanan Tanawuttiwat, JACC EP) was written on the possible promising new apporach of DGM.

Ventricular tachycardia · VT

We have tested DGM on mapped VTs using CARTO (Biosense Webster). Given the similarity in algorithms applicable to both ATs and VTs, our findings reveal that DGM is a rapid, automated algorithm exhibiting a strong agreement with traditional mapping for manually re-annotated VT maps. This collaborative research involved Associate Professor and Electrophysiologist Geoff Lee, along with Dr. and Electrophysiology Fellow Josh Hawson from the Royal Melbourne Hospital. The results of this study, utilizing DGM and thus focusing only on true reentry loops, have been published in JACC: Clinical Electrophysiology6.

Torsade de Pointes · TdP

The initial application of DGM (primitive version) focused on analyzing intramural needle data from the CAVB dog model developed in Prof. M. Vos's lab at Utrecht Medical Center. The primary question was whether TdP is perpetuated by focal activity or reentry. Preliminary analysis using an early version of DGM revealed that short TdP episodes are sustained by focal activity, while longer episodes involve reentry. Non-terminating TdPs consistently showed reentry. The findings of this research have been published in JACC: Clinical Electrophysiology.1.

After the results were published, an editoral was written on our work by Sachin Nayyar, Andreu Porta-Sánchez, and Kumaraswamy Nanthakumar (JACC EP). The question was raised how the reentry exactly manifests itself. Is it functional or anatomical? Are there multiple reentries? Is there meandering of the spiral core. This editorial inspired a follow-up paper where we sought to address some of these questions.

In our follow-up study7, we used the more mature DGM software, which confirmed our previous results. In addition, we found that non-terminating (NT) Torsade de Pointes (TdP) episodes consistently exhibited more simultaneous reentry loops compared to self-terminating (ST) episodes. (Bi-)ventricular loops were prevalent in non-terminating episodes. Thus we found that macro-reentry and multiple simultaneous localized reentries play a role in prolonged TdP. We also found that focal sources (which initiate each episode of TdP) tended to occur in preferred locations, suggesting potential implications for targeted treatment.

Atrial fibrillation · AF

We analyzed a simulated basket catheter data of a meandering spiral wave in the right atrium. When this data was analyzed with phase mapping, many false rotors appeared due to interpolation and far field effects (Martinez-Mateu et al., PLoS Comput Biol, 2018). By changing the parameters of DGM, we were able to differentiate the false rotors from the true rotors. Our manuscript on this study is published in Computers in Biology and Medicine4. Although a meandering spiral is not representative of AF, this was our first step in the direction of non-regular data.

We are currently analyzing simulated datasets of AF. To better understand AF is one of our main goals in the future.

Ventricular fibrillation · VF

We are currently testing DGM and comparing DGM with phase mapping on optical mapping datasets of VF in rats in collaboration with Prof. Dr. Fu Siong (Imperial College London).

DGM and Phase mapping

In our first study, we have compared DGM with phase mapping for a simulated stable rotor (functional reentry) whereby we put an regular electrode system on top of the rotor. We systematically added Gaussian noise to the LATs, and compares DGM with phase mapping. We found that phase mapping is very sensitive to noise, while DGM is much more robust.

In collaboration with Prof. Saiz and Dr. Martínez Mateu we have also compared phase mapping and DGM for the analysis of simulated basket catheter data in the right atrium for a meandering spiral. We found that DGM overcomes some of the limitations of phase mapping, which often finds false rotors4.

Finally, we conducted a comparative analysis of various phase mapping techniques alongside DGM for rotor detection, evaluating performance using 64 simulated 2D rotors with varying levels of fibrotic tissue, temporal noise, and meandering. Our study indicates that under conditions of low meandering, fibrosis, and noise, both PM and DGM yield comparable, excellent results. However, under conditions of high meandering, fibrosis, and noise, phase mapping is prone to errors, particularly an excess of false positives, leading to lower precision. In contrast, DGM demonstrates greater robustness in these scenarios. Our manuscript is currently under revision.

DGM challenge

DGM is a versatile tool capable of analyzing simulated, experimental, and clinical datasets. While phase mapping is commonly employed to identify rotors (functional reentry), it may not be optimal for detecting macro-reentry. DGM, on the other hand, excels in automatically analyzing various reentry types, including macro-reentry, rotors and focal sources. To facilitate comparison, we have integrated several versions of phase mapping into our DGM framework. This flexibility positions DGM as an excellent tool for analyzing diverse datasets. Another advantage lies in DGM's compatibility with any input type: 2D grid, 3D data, regularly or irregularly spaced data, and more. DGM only requires the input signals (intra-cardiac) or the LATs and the coordinates of the input electrodes. We are always eager to explore new collaborations to expand the applications of DGM, so contact us if you want to analyze an interesting dataset!

Publications