Abstract:

The most important ECG marker for the diagnosis of ischemia or infarction is a change in the ST segment. Baseline wander is a typical artifact that corrupts the recorded ECG and can hinder the correct diagnosis of such diseases. For the purpose of finding the best suited filter for the removal of baseline wander, the ground truth about the ST change prior to the corrupting artifact and the subsequent filtering process is needed. In order to create the desired reference, we used a large simulation study that allowed us to represent the ischemic heart at a multiscale level from the cardiac myocyte to the surface ECG. We also created a realistic model of baseline wander to evaluate five filtering techniques commonly used in literature. In the simulation study, we included a total of 5.5 million signals coming from 765 electrophysiological setups. We found that the best performing method was the wavelet-based baseline cancellation. However, for medical applications, the Butterworth high-pass filter is the better choice because it is computationally cheap and almost as accurate. Even though all methods modify the ST segment up to some extent, they were all proved to be better than leaving baseline wander unfiltered.

Abstract:

A common treatment of focal ventricular tachycardia is the catheter ablation of triggering sites. They have to be found manually by the physician during an intervention in a catheter lab. Thus, a method for determining the position of the focus automatically is desired. The inverse problem of electrocardiography addresses this problem by reconstructing the source of the ectopic beats using the surface ECG. This problem is ill-posed and therefore needs specific methods for solving it. We propose a machine learning approach for localisation of the ectopic foci in the heart to assist cardiologists with their therapy planning.We simulated 600 120-lead ECGs with different known excitation origins in the heart using a cellular automaton followed by a forward calculation. Features from the ECGs were used as input for a support vector regression (SVR). We assumed a functional relation between features from the ECG and the excitation origin. To benchmark SVR, we also used the well-known Tikhonov 0th order regularisation to reconstruct the transmembrane potentials in the heart and detect the location of the ectopic foci. Parameters for SVR and regularisation were chosen using a grid search minimising the error between estimated and true excitation origin. Compared to the Tikhonov regularisation method, SVR achieved a smaller deviation between estimated and real excitation origin evaluated with 6-fold cross validation. Future work could investigate on the behaviour on data from simulations with other torso and electrophysiological models, the influence of other methods for feature extraction and finally the evaluation with clinical data.

Abstract:

Chronic kidney disease appears worldwide. In the United States, the number of patients suffering from kid- ney failure doubled from 1998 to 2010. A common treat- ment for these patients is haemodialysis. However, the frequency of deaths caused by cardiovascular diseases is up to 10% to 30% higher in patients undergoing dialysis than in the general population. To analyse the underly- ing effects and for a possible risk prediction, a continuous monitoring of the ionic concentrations that are influenced by dialysis is desired. In this work, a method for the re- construction of the ionic concentrations of calcium and potassium from the ECG is proposed. In a first step, 91 monodomain simulations with the ten Tusscher ventricular cell model were performed for different extracellular ionic concentrations. From there, a standard 12-lead ECG was extracted. Calcium and potassium changes yielded ECGs clearly differing in amplitude and morphology. In a second step, the simulated ECG signals were used for reconstruc- ting the ionic concentrations directly from the ECG. Fea- tures were extracted from the signals designed to describe changes caused by varied ionic concentrations. The in- verse problem, i.e. coming back from the ECG features to the ionic concentrations was solved by regression with an artificial neural network. Results for potassium estimation yield an error of 0.00±0.28 mmol/l (mean±standard de- viation) calculated with 7-fold cross validation. The esti- mation error for calcium was 0.00±0.08 mmol/l. Although these results underline the suitability of the method, the used ECGs differed from the observed in a clinical envi- ronment. However, simulations allow an evaluation un- der controlled conditions of a particular effect that was intended to be investigated. As the application to clinical data is yet missing, this study can be seen as a proof of concept showing that an artificial neural network is capa- ble of exactly estimating potassium and calcium concen- trations from ECG features. 1. Introduction Haemodialysis therapy is a common treatment method for patients suffering from chronic kidney disease (CKD) in the terminal stage. The amount of people in the United States suffering from kidney failure increased from 320,000 in 1998 to 650,000 in 2010. The frequency of deaths caused by cardiovascular events within the dialysis patient group is up to 10% to 30% higher than in gene- ral population [1]. Patients suffering from end-stage CKD experience high variations of blood electrolyte concentra- tions. These can directly influence the functioning of the heart. Thus, research on cardiovascular links could im- prove therapy and risk stratification. One tool which is capable of capturing the electrophysiological properties of the heart in a non-invasive way is the electrocardiogram (ECG). It is known, that electrolyte concentrations of po- tassium (K+) and calcium (Ca2+) affect the ECG [2]. Un- til now, a determination of the concentrations is connec- ted to a blood test. Hence, continuous monitoring of the ionic concentration is impracticable. However, the ECG as a continuous, non-invasive monitoring tool could shed a light on the relation between heart diseases and changes in the ionic concentration particularly after leaving the strictly supervised clinical area where dialysis takes place, i.e allowing a monitoring at home. Articles have been pub- lished showing that the reconstruction of extracellular K+ concentration can be done using just one feature from the ECG with a quadratic regression [3]. In this study, we tried to estimate both K+ and Ca2+ concentrations from the ECG. Therefore, we examined simulated ECGs at dif- ferent concentration levels and designed features descri- bing the observed changes in the ECG. A subset of these was used in connection with a machine learning method to reconstruct the concentrations. 2. Methods 2.1. Simulations A total number of 91 computer simulations of the car- diac electrophysiology were performed at whole heart

Abstract:

Atrial arrhythmias like atrial fibrillation and atrial flutter are a major health challenge in developed countries. Radiofrequency ablation performed via intracardiac catheters is a curative therapy for these reentrant arrhythmias. However, the optimal location of ablation lesions is not straightforward to determine, particularly for complex activation patterns. Thus, a clinical need for tools to intuitively visualize complex activation patterns and to provide a platform to evaluate different ablation strategies in dry runs is apparent. Here, we present a virtual reality system that allows to interactively simulate atrial excitation propagation and place ablation lesions. Our software builds on the IMHOTEP framework for the Unity3D engine and implements a multithreaded model-view-controller design pattern. Excitation propagation is computed using a fast marching approach considering refractoriness. Interactive rewind and playback is supported through a combination of the flyweight pattern for simulation data with complete snapshots for key frames. The system was evaluated in a user study using the HTC ViveTM headset including two controllers. For high fidelity virtual reality interaction, a minimum frame rate of 60 per second is required. In a biatrial anatomical model comprising 36,059 nodes (Figure 1), even complex activation patterns with multiple wavefronts could be simulated and rendered down to 2x slow motion (1 sec activation sequence displayed during 2 sec wall time) on a desktop machine. Results of the user study suggest added value regarding the comprehension of arrhythmias and ablation options and very good intuitiveness of the user interface requiring almost no teach-in. The virtual reality tool is ready to be used for educational purposes and prepared to import personalized models supporting diagnosis and therapy planning for atrial arrhythmias in the future.