Validation of a Novel Catheter Guiding Method for the Ablative Therapy of Ventricular Tachycardia in a Phantom Model

Abstract

Accurate guidance of an ablation catheter is critical in the radio-frequency ablation of ventricular tachycardia. With current technologies it is challenging to rapidly and accurately localize the site of origin of an arrhythmia, often restricting treatment to patients with hemodynamically stable arrhythmias. We investigated the effectiveness of a new guidance method, the Inverse Solution Guidance Algorithm (ISGA), which is based on a single-equivalent dipole representation of cardiac electrical activity and is suitable for patients with hemodynamically unstable VT. Imaging was performed in homogeneous and inhomogeneous saline-filled torso phantoms in which a catheter tip was guided towards a stationary electrical dipole source over distances of more than 5 cm. Using ISGA, the moving catheter tip was guided to within 0.61±0.43 mm and 0.55±0.39 mm of the stationary source in the homogeneous and inhomogeneous phantoms, respectively. This accuracy was achieved with less than 10 movements of the catheter. These results suggest that ISGA has potential to provide accurate and efficient guidance for radio-frequency ablation procedures in the patient population with hemodynamically unstable arrhythmias.

I. Introduction

More than 200,000 deaths a year in the United States may be attributable to ventricular tachycardia (VT) post-myocardial infarction [1]. Radio-Frequency Ablation (RFA) of the site of origin of the arrhythmia may permanently abolish the VT and is therefore currently the optimal treatment [2, 3]. We have previously described a method to guide the electrophysiologist to the optimal site for ablation, the Inverse Solution Guidance Algorithm (ISGA). This algorithm utilizes a single equivalent moving dipole (SEMD) model of electrical activity to localize the exit site of the re-entrant circuit [4–7]. To accurately guide an ablation catheter to the exit site, ISGA then estimates the location of the ablation catheter tip by calculating the location and orientation of a current dipole generated between two electrodes at its tip [5]. Real space contains the real catheter tip and real arrhythmogenic dipole; image space contains the dipole solutions for the electrical activity at these two locations estimated using ISGA. By manipulating the catheter so as to bring the dipole positions together in image space, the ablation catheter tip may be guided to the exit site of the re-entrant circuit for the accurate delivery of ablative energy. The localized electrical activity at the exit site of the re-entrant circuit shall be referred to as the arrhythmogenic dipole. The dipole solutions for electrical activity at the catheter tip and exit site shall be referred to as the catheter dipole image and arrhythmogenic dipole image. Since a computationally-simple infinite, homogeneous model is used to estimate the potentials on the torso surface due to a dipole source in real-time, the arrhythmogenic dipole image is displaced from the true position of the arrhythmogenic dipole [8]. Previous studies have indicated the potential of ISGA to co-localize two electrical sources and thus to be used as a guidance method [4, 5, 7, 9]. However, experimental measurements with inhomogeneous phantom models have not been conducted.

In this study, we performed the first investigation of ISGA’s accuracy and efficiency as a guidance algorithm in an inhomogeneous phantom. Three measures were used to evaluate its success. First, the accuracy with which the moving catheter tip could be superposed with the stationary dipole in both homogeneous and inhomogeneous phantom models was calculated. Second, a quantitative comparison was made of the distances and directions moved in real and image space. Third, the magnitude of the displacement between a dipole’s position in real and image space was calculated.

II. METHODS

A. Phantom Model and Signal Measurement

The phantom model consisted of an open cylindrical tank of radius 14.6 cm, filled with standard saline solution to a height of 46 cm to approximate the dimensions of an average male torso (see Figure 1). 60 Ag-AgCl electrodes were arrayed over the vertical surface of the phantom in a rhombic lattice configuration. One electrode at the base of the phantom was chosen as the reference electrode. The phantom and the proximally unshielded electrode leads were enclosed within a Faraday cage to minimize 60 Hz noise. A rigid catheter with two platinum electrodes 3 mm apart at its tip was vertically mounted on a 3-D positioning system and placed inside the phantom. The positioning system allowed the catheter tip location to be measured with an accuracy of 0.2 mm in real space and moved within a 20-cm sided ‘heart volume’ approximately centered within the phantom. A sinusoidal signal of frequency 100 Hz and amplitude 10 V was generated between the catheter tip electrodes by an electrically isolated BK Precision model 4011A function generator. This created potentials ranging from 1 to 10 mV at the surface electrodes, approximating the amplitude of a surface ECG. A sinusoidal signal was chosen instead of a VT waveform for computational simplicity. It should be noted that only instantaneous signal amplitudes are used by the Inverse Solution Guidance Algorithm, as previously described [4–6]. Therefore it was not necessary to replicate the frequency components of an ECG signal.

Fig. 1.

Experimental Setup, showing the phantom model, the electrode layout, and an example configuration of the three objects used to create an electrically inhomogeneous phantom (these were of dimensions r=7.5 cm, h=37 cm; r=9 cm, h = 28 cm; and r = 8.3 cm, h=8 cm, and were arranged in a different configuration for each trial.

The differential signals between the 59 measurement electrodes and 1 reference electrode were amplified using WPI ISO-DAM8 isolated bioamplifiers. Electrical isolation of the amplifiers ensured that current leakage through the surface electrodes was insignificant compared to the amplitude of the current between the catheter tip electrodes. The data in the 59 channels was sampled at 1 kHz and filtered with a 5^th order Butterworth band-pass filter to retrieve the sinusoidal signal amplitudes.

A graphical user interface enabled real-time assessment of signal quality from each channel. It also displayed the estimated locations and orientations of the stationary target dipole and moving catheter tip dipole. A dipole’s location and orientation were represented using a ball and arrow, respectively. The distance and vector between the two dipole images (the inter-image vector) were displayed numerically.

B. Inverse Solution Guidance Algorithm

The Inverse Solution Guidance Algorithm has been described previously [4–6]. In the application of the Inverse Solution Guidance Algorithm, voltages created by a dipole within the phantom are measured at the multiple electrodes on the phantom surface. A χ² measure of the goodness-of-fit between measured and modeled surface voltages is then used to calculate the dipole position and orientation; a brute-force search of the χ² objective function space (that avoids local minima) is then conducted within the volume conductor, and the dipole whose parameters minimize the objective function is the solution to the inverse problem. Due to the proximity of the reference electrode to the measurement electrodes in the phantom model, the accuracy of the SEMD solution was improved by taking into account the significant 100 Hz signal at the reference electrode. Therefore, the forward model describing the estimated potential, ϕⁱ, in the i^th channel due to a single dipole in an infinite, homogeneous volume conductor was adapted:

$${\phi }^i=\frac{\mathbf{p}·(\mathbf{r}-{\mathbf{r}}_i\text{'})}{4\pi g{|\mathbf{r}-{\mathbf{r}}_i\text{'}|}^3}-\frac{\mathbf{p}·(\mathbf{r}-{\mathbf{r}}_{\mathit{\text{ref}}}\text{'})}{4\pi g{|\mathbf{r}-{\mathbf{r}}_{\mathit{\text{ref}}}\text{'}|}^3}$$

where r_ref represents the reference electrode location.

C. Experimental Protocol

The catheter was placed at a random position within the phantom and its 3-D coordinates in real space (the stationary dipole location) and in image space (the stationary dipole image location) noted. The catheter was then moved in all three dimensions by a minimum of 5 cm to a second random position. The catheter’s new 3-D coordinates in real space (the initial moving dipole location) and image space (the initial moving dipole image location) were recorded. The interface simultaneously displayed the stationary and moving dipole images. As second operator then attempted to guide the catheter tip to the location of the stationary dipole using only the information displayed on the user interface, and moving the catheter tip each time in real space by between one-half and two-thirds of the remaining distance between the images. Superposition of the dipole images was indicated by a distance between them of less than 0.5 mm on the graphical interface. Since the same catheter was used to generate both the stationary and moving dipoles, there was no collision feedback to the operator if the moving dipole was brought to the location of the stationary dipole inside the phantom.

15 trials were conducted in a homogeneous phantom, each time using a randomly-chosen stationary dipole location and randomly-chosen initial moving dipole location. A further 15 trials were conducted in an inhomogeneous phantom. Torso inhomogeneities were simulated using three cylindrical plastic objects (of dimensions r=7.5 cm, h=37 cm; r=9 cm, h = 28 cm; and r = 8.3 cm, h=8 cm) placed in different configurations within the phantom (see Figure 1). The objects were arranged at the start of each trial in a random configuration that would not obstruct the movement of the catheter tip towards the stationary dipole.

III. RESULTS

A. Accuracy of Dipole Superposition

The catheter tip was guided to within 0.61± 0.43 mm and 0.55 ± 0.39 mm of the stationary dipole in the homogeneous and inhomogeneous phantom studies, respectively. This difference in guidance accuracy is not significantly different at the 0.05 level, indicating that guidance accuracy does not change in the presence of electrical inhomogeneities. A single outlier with an end-point accuracy of 28.2 mm is excluded from our stated end-point accuracy for the inhomogeneous phantom. For this outlier, unlike the other data points, it was observed that when the dipole images were superposed, their orientations were not aligned. Furthermore, when the catheter tip was moved to the real location of the stationary dipole, the dipole image locations and orientations were both superposed. It was ascertained that the outlier result was not in fact a result of a non-linear parameter-fitting error, in which the algorithm had become caught in a local minimum of the χ² goodness-of-fit objective function. In fact, our previously-described brute force search method correctly found the absolute χ² minimum within the volume conductor, as it is designed to do [10]. Instead, because of severe distortions in image space caused by the electrical inhomogeneities, two dipoles at different locations in real space mapped to dipoles at the same location but different orientations in image space.

B. Comparison of Real and Image Space

The distances of the catheter-tip from the arrhythmogenic dipole in image space (i_d) and real space (r_d) were compared over multiple trials for m positions of the catheter-tip, in the homogeneous (m = 95) and inhomogeneous (m = 119) phantom models. The results are shown in Figure 2 A and B. The linear relationship of i_d to r_d shown in the figures indicates that distances in real and image space are almost equal even in the presence of significant non-idealities (the scattering of data points about the best-fit polynomial is more limited in the homogeneous case). The number of movements of the catheter required to superpose the images of the moving and stationary dipoles was also small – 7.33 ± 2.09 and 7.81 ± 2.14 for the homogeneous and inhomogeneous trials, respectively (the difference is not statistically significant). The difference between the angles made in real vs. image space was 5.38 ± 1.94 degrees and 4.65 ± 1.76 degrees for the homogenous and inhomogeneous phantoms, respectively.

Figs. 2.

A and B: Comparison of the distance of the catheter-tip from the arrhythmogenic dipole in image space (i_d) and real space (r_d) for m positions of the catheter-tip as it moves towards the arrhythmogenic dipole, in a). the homogeneous (m = 95) and b). inhomogeneous (m = 119) phantom models respectively. The least-squares first-order polynomial fit of i_d vs. r_d (black solid line) whose function is shown with each figure, indicates that the ratio of i_d to r_d is approximately 1 in both cases, while the 95% confidence intervals (red dash-dotted line) indicate a good quality of fit.

IV. DISCUSSION

In this study, the accuracy with which our Inverse Solution Guidance Algorithm could be used to guide a catheter tip to the site of a co-oriented stationary dipole was investigated in homogeneous and inhomogeneous phantom models. An accuracy of significantly less than 1 mm was attained in all 15 experiments in the homogeneous phantom and in 14 out of 15 experiments in the inhomogeneous phantom. Furthermore, our results indicate that movements of the catheter tip dipole in real and image space correspond both in distance and direction even when significant inhomogeneities and boundary effects are present in the volume conductor.

The single outlier in the inhomogeneous phantom trials indicates that the Inverse Solution Guidance Algorithm may falsely indicate the spatial superposition of two dipoles. However, for this outlier, while the dipoles were spatially superposed in image space, their orientations were different. Therefore, in the presence of significant non-idealities and sources of systematic error, dipoles at multiple locations in real space may map to dipoles with the same location in image space but different orientations. Therefore, a method to compensate for the effect of dipole orientation in the presence of sources of systematic error is necessary. An algorithm utilizing a catheter design with multiple tip electrodes, arranged such that three independent dipoles may be generated between them, has been developed. By linearly-superposing the three independent dipoles, a catheter tip dipole may be created whose orientation matches that of the target at each step. This algorithm has been tested in simulations and has been found to improve convergence in the presence of sources of systematic error [10]. Future work will involve development of such a catheter and testing of the method in a phantom torso.

There are several limitations to this study. Firstly, the human torso contains electrical anisotropies which have not been simulated in this study. In addition, it is likely that in an animal or human, exact electrode positions would not be known. However, these are both simply additional forms of systematic error and it is therefore expected that the proposed algorithm would work nevertheless. While the phantom model used in this study does not have a realistic torso shape, we would expect our results to improve if a more realistic geometry were used; surface electrodes on the front of the chest are closer to the heart than those in our phantom model, and the convex surface of a real torso would improve the algorithm’s accuracy along the torso axis. It should be noted however that in a real system, sources of systematic error will not be constant. Breathing and cardiac movement cause variations in electrode positions and in the distribution of torso inhomogeneities. To overcome this, the use of controlled breath-holding and image-gating by surface ECG signals would need to be explored in a clinical setting.

We have tested the Inverse Solution Guidance Algorithm in a phantom model and shown that it provides an accurate and efficient method for the guidance of a catheter tip to the location of a stationary dipole. Since this method is able to image the site of origin of the arrhythmia from a single beat of the tachycardia, ISGA is suitable for use even in patients with hemodynamically unstable VT. Therefore, ISGA may allow RFA to be performed not only more accurately, but also in a much wider segment of the population affected by VT than is currently possible.