Journal publications
Noninvasive reconstruction of cardiac electrical activity: update on current methods, applications and challenges.
Cluitmans, M. J. M.; Peeters, R. L. M.; Westra, R. L. & Volders, P. G. A.
Neth Heart J, 2015, 23, 301-311
Abstract: Electrical activity at the level of the heart muscle can be noninvasively reconstructed from body-surface electrocardiograms (ECGs) and patient-specific torso-heart geometry. This modality, coined electrocardiographic imaging, could fill the gap between the noninvasive (low-resolution) 12-lead ECG and invasive (high-resolution) electrophysiology studies. Much progress has been made to establish electrocardiographic imaging, and clinical studies appear with increasing frequency. However, many assumptions and model choices are involved in its execution, and only limited validation has been performed. In this article, we will discuss the technical details, clinical applications and current limitations of commonly used methods in electrocardiographic imaging. It is important for clinicians to realise the influence of certain assumptions and model choices for correct and careful interpretation of the results. This, in combination with more extensive validation, will allow for exploitation of the full potential of noninvasive electrocardiographic imaging as a powerful clinical tool to expedite diagnosis, guide therapy and improve risk stratification.
Conference proceedings
In-vivo Evaluation of Reduced-Lead-Systems in Noninvasive Reconstruction and Localization of Cardiac Electrical Activity
Cluitmans, M. J.; Karel, J.; Bonizzi, P.; de Jong, M. M.; Volders, P. G.; Peeters, R. L. & Westra, R. L.
Computing in Cardiology, to be published, 2015
Abstract: Noninvasive imaging of electrical activity of the heart has increasingly gained attention over the last decades. Epicardial potentials can be reconstructed from a torso-heart geometry and body-surface potentials recorded from tens to hundreds of body-surface electrodes. However, it remains an open question how many body-surface electrodes are needed to accurately reconstruct epicardial potentials.
In a canine model, we reconstructed epicardial electrograms from 39 beats, investigating the use of a full-lead system, consisting of 169 well connected body-surface electrodes, and reduced-lead systems: using half or a third of the electrodes, or a minimalistic set of the default 12-lead ECG (10 or 9 electrodes).
Invasively recorded epicardial electrograms were used to evaluate the quality of the reconstructed electrograms. In 27 paced beats, also the mismatch between the known pacing location and reconstructed location of earliest activation was compared.
Correlation coefficients indicate that the quality of the reconstructed electrograms remains stable from 169 (full set) to 59 (a third) electrodes, and decreases with fewer electrodes. The mismatch between the detected origin of a beat and known pacing location increases with fewer electrodes, see Figure. However, it is remarkable that with only 9 or 10 electrodes, the median mismatch between detected and known pacing location is 30mm, only marginally higher than when half of the electrodes is used.
This study indicates that for specific goal such as detecting the origin of an extrasystolic beat, a lower number of body-surface electrodes can provide noninvasive reconstructions that can be still useful in clinical practice. However, especially when using only a 12-lead setup as suggested by others, it is important to realize that although the median results are reasonable, there is a large spread of the error, up to a 65mm mismatch.
Physiology-based Regularization Improves Noninvasive Reconstruction and Localization of Cardiac Electrical Activity
Cluitmans, M. J.; de Jong, M. M.; Volders, P. G.; Peeters, R. L. & Westra, R. L.
Computing in Cardiology, 2014, 41, 1-4
Abstract: The objective of the inverse problem of electrocardiography is to noninvasively reconstruct information about electrical activity at the heart surface (epicardium), from electrical measurements on the body surface and a patient-specific torso-heart geometry. This is complicated by the ill-posedness of the inverse problem, making reconstructions imperfect. Previously, we have shown that a realistic basis can be created from (simulated) epicardial training potentials. Reconstructions from traditional methods can be projected onto this basis, improving the quality of reconstructions. Here, we propose a novel superior method called ‘physiology-based regularization’ that renders traditional (e.g. Tikhonov) reconstruction and projection unnecessary. Instead, reconstruction of epicardial electrograms is achieved directly, by pursuing a sparse representation in terms of this realistic basis. This approach differs radically from the traditional regularization methods, which apply mathematical or physical constraints. By using a realistic epicardial basis, it is possible to include physiological knowledge as constraint. We validate this method by simultaneously recording 67 invasive epicardial electrograms in a canine experiment. The correlation between physiology-based reconstructed and in vivo measured electrograms (Pearson’s r = 0.60) is similar to those of traditional reconstruction methods such as Tikhonov regularization (r=0.59). We further demonstrate that by creating a realistic basis for a specific pathology, this method can answer clinical questions with improved accuracy.
Ultimately, physiology-based regularization would improve patient care by yielding patient-specific results, inspired by electrophysiological knowledge and optimized to answer clinically relevant questions.
Wavelet-sparsity based Regularization over Time in the Inverse Problem of Electrocardiography
Cluitmans, M.; Karel, J.; Bonizzi, P.; Volders, P.; Westra, R. & Peeters, R.
Conf Proc IEEE Eng Med Biol Soc., 2013, 3781-3784
Abstract: Noninvasive, detailed assessment of electrical cardiac activity at the level of the heart surface has the potential to revolutionize diagnostics and therapy of cardiac pathologies. Due to the requirement of noninvasiveness, body-surface potentials are measured and have to be projected back to the heart surface, yielding an ill-posed inverse problem. Ill-posedness ensures that there are non-unique solutions to this problem, resulting in a problem of choice. In the current paper, it is proposed to restrict this choice by requiring that the time series of reconstructed heart-surface potentials is sparse in the wavelet domain.
A local search technique is introduced that pursues a sparse solution, using an orthogonal wavelet transform. Epicardial potentials reconstructed from this method are compared to those from existing methods, and validated with actual intracardiac recordings. The new technique improves the reconstructions in terms of smoothness and recovers physiologically meaningful details. Additionally, reconstruction of activation timing seems to be improved when pursuing sparsity of the reconstructed signals in the wavelet domain.
Inverse Reconstruction of Epicardial Potentials Improves by Vectorcardiography and Realistic Potentials
Cluitmans, M. J.; Bonizzi, P.; Karel, J. M.; Volders, P. G.; Peeters, R. L. & Westra, R. L.
Computing in Cardiology, 2013, 40, 369-372
Abstract: The inverse problem of electrocardiography is to reconstruct electrical activity at the level of the epicardium from body-surface electrograms and a patient-specific torso-heart geometry. Previously, we have shown that the use of realistic epicardial training electrograms can improve reconstruction quality in silico. Here, we apply this method in a patient and compare the resulting computed electrograms with selected intracardiac measurements. Additionally, we utilize a new method that yields further improvements by incorporating characteristics of vectorcardiographic patient-specific information.
In a patient, 256 body-surface electrograms were recorded simultaneously. A CT scan was performed to obtain the position of electrodes and the heart contour. Based on the digitized torso-heart geometry and measured body-surface potentials, Tikhonov regularization was applied to compute the associated epicardial electrograms. As a novel step, these electrograms were then improved by optimizing their vectorcardiographic characteristics to match those from body-surface potentials by means of subspace alignment. Next, improved epicardial electrograms were projected on a realistic basis, constructed from computer-generated (FitzHugh-Nagumo) electrograms.
Two intracardiac leads provided measurements: at the left ventricular epicardium, and the right ventricular endocardium. The measured electrograms were compared to the reconstructed electrograms at those locations, for a paced beat and a native beat. Tikhonov reconstruction alone yielded reconstructed electrograms with a Spearman correlation coefficient (CC) of the QRS-complexes of 0.03. Vectorcardiographic improvement yielded electrograms with CC=0.64, and subsequent projection onto the realistic basis resulted in electrograms with CC=0.77, which was a significant improvement. More importantly, as the figure illustrates for a paced beat on the left ventricle, significant deflections showed up in the reconstructed electrograms after applying vectorcardiographic improvement and projection on the realistic basis (“Tikh+VCG+RealBasis”).
These results indicate that computer-generated electrograms are useful for improving reconstruction of epicardial electrograms, and that the combined use of patient-specific vectorcardiographic characteristics yields further improvements.
Realistic Training Data Improve Noninvasive Reconstruction of Heart-Surface Potentials
Cluitmans, M.; Peeters, R.; Volders, P. & Westra, R.
Conf Proc IEEE Eng Med Biol Soc., 2012, 6373-6376
Abstract: The inverse problem of electrocardiography is to noninvasively reconstruct electrical heart activity from body-surface electrocardiograms. Solving this problem is beneficial to clinical practice. However, reconstructions cannot be obtained straightforwardly due to the ill-posed nature of this problem. Therefore, regularization schemes are necessary to arrive at realistic solutions. To date, no electrophysiological data have been used in reconstruction methods and regularization schemes. In this study, we used a training set of simulated heart-surface potentials to create a realistic basis for reconstructions of electrical cardiac activity. We tested this method in computer simulations and in one patient. The quality of reconstruction improved significantly after projection of the results of traditional regularization methods on this new basis, both in silico (p0.01) and in vivo (p0.05). Thus, we demonstrate that the novel concept of applying electrophysiological data might be useful to improve noninvasive reconstruction of electrical heart activity.
Abstracts
Whole-heart simulations improve localization of origin of a heart beat in noninvasive electrocardiographic imaging
Cluitmans, M.; Clerx, M.; Karel, J.; Bonizzi, P.; Peeters, R.; Volders, P. & Westra, R.
International Conference on Basic and Clinical Multimodal Imaging, 2015
Abstract: Whole-heart simulations improve localization of origin of a heart beat in noninvasive electrocardiographic imaging
Noninvasive electrocardiographic imaging can reconstruct electrical potentials, electrograms, and isochrones on a patient-specific heart geometry. By localizing the origin of extrasystolic beats, this modality can expedite catheter ablation therapy and improve patient care. However, electrocardiographic imaging suffers from numerical instability due to ill-posedness, requiring the use of regularization methods to stabilize the solution.
Recently, we have introduced a new regularization method, which aims at incorporating physiological knowledge instead of physical constraints. We have shown that a non-cardiac action potential (AP) model can be used on a patient-specific heart geometry to generate potential patterns that can be used as a basis for noninvasive reconstruction of cardiac potentials at the epicardium. In the present study, we have refined this approach by using an AP model specific to ventricular cardiac tissue, namely the 1991 Luo-Rudy model. This was achieved by using Myokit, a toolbox for AP simulations.
This approach was evaluated in a canine experiment. Pacing was performed at 24 different epicardial locations in a healthy dog. A physiological basis was generated by simulating 32 beats originating from randomly selected locations on the dog’s digitized epicardial surface. Additionally, physiological bases were generated with beats only originating from the left or right side of the heart, respectively. Epicardial potentials were reconstructed with the general physiological basis, and with the ‘side-specific’ bases for beats originating from that side of the heart. Activation times were determined as the moment of steepest downslope in reconstructed electrograms. Localization mismatch was determined as the distance between the known pacing location and the reconstructed location of earliest activation.
Commonly used Tikhonov zeroth order regularization resulted in a mean localization mismatch of 10±5 mm, while physiology-based regularization with a general basis yielded a mismatch of 16±11 mm and the side-specific bases a mismatch of 11±4 mm (left) and 12±5 mm (right).
This shows that physiology-based regularization currently performs as well as traditional regularization methods. Additionally, it can be tuned to a specific purpose by using patient-specific whole heart modelling. We expect that results can be improved further by the inclusion of more patient-specific information, for example, genetic mutations or areas of scar tissue (based on MRI or other imaging).
High-resolution validation of localization of electrocardiographic imaging shows a centimeter-accuracy
Cluitmans, M.; de Jong, M.; Karel, J.; Bonizzi, P.; Prinzen, F.; Peeters, R.; Westra, R. & Volders, P.
International Society for Computerized Electrocardiology (ISCE) 40th Annual Conference, San Jose, California, USA, 15-19 April 2015, 2015
Abstract: Electrocardiographic imaging allows for noninvasive reconstruction of epicardial electrograms, activation isochrones, repolarization times and other electrical quantities of the heart. Despite its increasing clinical use, in vivo validation has been performed only to a limited amount. Little is known about the accuracy of reconstructions of available set-ups; claims exist of accuracy of 10 mm or less. We present the results of a canine validation experiment, for the first time assessing the accuracy that can be achieved by electrocardiographic imaging with high-precision, simultaneously acquired validation data.
In a healthy dog, 99 electrodes were implanted around the epicardium, 192 body-surface electrodes were attached to the torso, and a torso-heart geometry was digitized from a high-resolution CT scan, providing the location of all electrodes with high accuracy. Potential recordings were obtained simultaneously at the epicardial surface and body surface, and pacing was performed on several epicardial locations. The electromagnetic relationship between the heart and body surface was based on the potential-based formulation of the inverse problem and was calculated on a homogeneous torso with publicly available methods. Epicardial electrograms were reconstructed with well-known zeroth order Tikhonov regularization and compared to the recorded epicardial electrograms.
Comparison of reconstructed with recorded epicardial electrograms on five beats (one sinus beat, four beats paced on different epicardial locations) shows that depolarization electrograms are more accurately reconstructed than repolarization electrograms (averaged Pearson’s correlation coefficient for all epicardial electrodes and all beats: QRS-segment 0.74 vs ST-segment 0.58, p0.05). Localization of earliest activation in 12 beats paced on different locations shows a pacing localization mismatch ranging from 4 to 57 mm with a median of 33 mm.
This experiment indicates that electrocardiographic imaging can noninvasively reconstruct activation electrograms with reasonably high correlation coefficients, and in a patient-specific anatomy. Repolarization abnormalities are important risk stratifiers for arrhythmias but repolarization electrograms are more difficult to reconstruct than activation characteristics. Moreover, localization of the origin of a beat comes with a considerable mismatch range, limiting the detection of origin of arrhythmias to a range of centimeters, not millimeters. Although clinical advantages are shown in an increasing number of studies, clinicians should be aware that accuracy of reconstructed signals may be limited.
Wavelet-sparsity based regularization
Cluitmans, M.; Karel, J.; Bonizzi, P.; Volders, P.; Peeters, R. & Westra, R.
ECG Imaging 2015 Workshop, Bad Herrenalb, Germany, 25-28 March 2015, 2015
Abstract: Noninvasive electrocardiographic imaging is performed in clinics with increasing frequency. However, due to the ill-posed character of the inverse problem, noninvasive reconstruction of electrical activity at the heart surface still suffers from suboptimal accuracy, particularly in terms of a high variance. We aim at improving accuracy of reconstructed epicardial electrograms to expedite diagnosis and guide therapy. We do so in the potential-based formulation of the inverse problem, using boundary-element methods in a completely homogeneous torso.
A novel approach is to pursue sparsity in the wavelet-representation of the reconstructed epicardial potentials. We investigate this method in a realistic torso-heart geometry with simulated data. Epicardial potentials were measured in a canine experiment. A transfer matrix provided the corresponding in silico body-surface potentials, to which we added noise to obtain a signal-to-noise ratio of 35dB. Epicardial potentials were then reconstructed based on the noisy body-surface potentials and the same transfer matrix. In an iterative approach, we pursued sparsity of the reconstructed epicardial potentials in the wavelet domain per epicardial node over time. For comparison, we also reconstructed epicardial potentials with traditional zeroth order Tikhonov regularization.
Results are shown in Figure. Clearly, Tikhonov reconstructed potentials suffer from noise in the reconstructions. On the other hand, wavelet-based sparsity results in electrograms that are very similar in terms of morphology to the original waveforms.
However, in in vivo data we have not yet been able to show significant improvements with wavelet-based regularization over traditional Tikhonov regularization. Nevertheless, this new approach has the potential to improve accuracy in a clinical setting as well, and future studies will focus on in vivo issues to achieve this improvement.
Noninvasive Imaging of Repolarization Gives Insight in Arrhythmogenic Substrate
Cluitmans, M. & Volders, P.
Nederlandse Vereniging voor Cardiologie (NVVC) Voorjaarscongres, Noordwijk, the Netherlands, 9-10 April 2015, 2015
Abstract: Purpose: Repolarization abnormalities are a well-known substrate for arrhythmias. However, subtle repolarization abnormalities are not easily interpreted from the 12-lead body-surface electrocardiogram. Noninvasive imaging of electrical activation and recovery directly at the epicardial surface might give insights in repolarization abnormalities.
Methods: Noninvasive electrocardiographic imaging was performed in a 47-year old female who presented with polymorphic ventricular tachycardia (VT) and frequent ventricular extrasystolic (VES) beats, without structural heart disease. Based on 184 simultaneously recorded body-surface ECGs and a torso-heart geometry (digitized from CT scan), local epicardial potentials, electrograms and isochrones were computed.
Results: Epicardial isochrones show normal ventricular activation during sinus rhythm (panel A), but abnormal recovery with steep repolarization gradients (55ms/cm; normal values reported: 2ms/cm). VES activation (B2,C2) follows the pattern of recovery (B1,C1) of the preceding sinus beat. Activation-recovery intervals (not shown) are similarly distributed as recovery times, indicating that repolarization gradients are induced by local changes in action potential duration (APD). It is likely that a short-coupled VES beat can initiate VT when late-recovering tissue is initially still refractory (D).
Conclusion: Noninvasive electrocardiographic imaging can help identify the mechanisms of VT by imaging recovery patterns and repolarization gradients. In this case a primary electrical disease is suspected to result in local APD heterogeneity and treatment with quinidine completely abolished extrasystolic activity.
Inverse reconstruction of epicardial potentials is improved by novel regularization methods and extensive validaton
Cluitmans, M. J.; Bonizzi, P.; Karel, J. M.; Peeters, R. L.; Volders, P. G. & Westra, R. L.
Mathematical Biosciences Institute (MBI) (Ed.)
Workshop 3: Integrating Modalities and Scales in Life Science Imaging, Columbus, Ohio, USA, 17-21 March 2014, 2014
Singular value decomposition and change of basis improve the inverse solution in electrocardiography
Cluitmans, M.; Westra, R. & Peeters, R.
3rd Dutch Conference on Bio-Medical Engineering (BME 2011), Egmond aan Zee, 20-21 Jan. 2011, 2011
The inverse solution of electrocardiography is improved by including localized, physiological knowledge
Cluitmans, M.; Westra, R. & Peeters, R.
30th Benelux Meeting on System and Control, Lommel, Belgium, 15-17 March 2011, Universiteit Gent, 2011, 197