Source localization of focal ventricular arrhythmias using linear estimation, correlation, and back propagation networks


Yilmaz B. , Cunedioglu U.

COMPUTERS IN BIOLOGY AND MEDICINE, cilt.37, ss.1437-1445, 2007 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 37 Konu: 10
  • Basım Tarihi: 2007
  • Doi Numarası: 10.1016/j.compbiomed.2007.01.008
  • Dergi Adı: COMPUTERS IN BIOLOGY AND MEDICINE
  • Sayfa Sayıları: ss.1437-1445

Özet

Catheter-based approaches used in the localization and treatment of the source of heart rhythm disturbances (arrhythmias) have become popular, because they do not require highly invasive and risky open-chest operations. In most of the existing approaches, mapping of the outer surface (epicardium) is not possible even though arrhythmic substrates involving epicardial and subepicardial layers account for about 15% of the ventricular tachycardias. In this study, we report a feasibility study of a novel mapping approach targeting the epicardium which is based on the measurements of multielectrode catheters placed in the coronary veins. We investigated three methods in determining the most probable region of early activation, i.e., the region that contains the source of the abnormal activation on the heart, using only a set of sparse venous catheter recordings. The methods we proposed here were the linear estimation, correlation, and the back propagation networks. The linear estimation technique hypothesized the relationship between venous catheter measurements and unmeasured epicardial sites based on a previously recorded training data set. The correlation method included a comparative analysis between test and training epicardial activation time maps based on the measured values from the venous sites. In the back propagation method, the input layer consisted of the source data in the form of 42 nodes which were the activation time values from the intravenous catheter leads. We used two hidden layers with 600 and 500 nodes, respectively. The output layer consisted of 28 nodes in the output layer that corresponded to the manually selected early activation regions on the epicardium. The results of the linear estimation and the correlation methods showed that they could be used as a good predictor for the region of early activation, and thus, these approaches may be employed to direct a subsequent more focused electrophysiological study and curative radiofrequency (RF) ablation. The back propagation network approach performed relatively well for the right ventricularly paced beats and the results demonstrated its potential as a supporting technique to the estimation and correlation methods. The results of this study encourage further investigation and provides evidence that an epicardial mapping approach based on the venous catheter recordings is feasible and can provide adequate accuracy for clinical applications. (c) 2007 Elsevier Ltd. All rights reserved.

Catheter-based approaches used in the localization and treatment of the source of heart rhythm disturbances (arrhythmias) have become popular, because they do not require highly invasive and risky open-chest operations. In most of the existing approaches, mapping of the outer surface (epicardium) is not possible even though arrhythmic substrates involving epicardial and subepicardial layers account for about 15% of the ventricular tachycardias. In this study, we report a feasibility study of a novel mapping approach targeting the epicardium which is based on the measurements of multielectrode catheters placed in the coronary veins. We investigated three methods in determining the most probable region of early activation, i.e., the region that contains the source of the abnormal activation on the heart, using only a set of sparse venous catheter recordings. The methods we proposed here were the linear estimation, correlation, and the back propagation networks. The linear estimation technique hypothesized the relationship between venous catheter measurements and unmeasured epicardial sites based on a previously recorded training data set. The correlation method included a comparative analysis between test and training epicardial activation time maps based on the measured values from the venous sites. In the back propagation method, the input layer consisted of the source data in the form of 42 nodes which were the activation time values from the intravenous catheter leads. We used two hidden layers with 600 and 500 nodes, respectively. The output layer consisted of 28 nodes in the output layer that corresponded to the manually selected early activation regions on the epicardium. The results of the linear estimation and the correlation methods showed that they could be used as a good predictor for the region of early activation, and thus, these approaches may be employed to direct a subsequent more focused electrophysiological study and curative radiofrequency (RF) ablation. The back propagation network approach performed relatively well for the right ventricularly paced beats and the results demonstrated its potential as a supporting technique to the estimation and correlation methods. The results of this study encourage further investigation and provides evidence that an epicardial mapping approach based on the venous catheter recordings is feasible and can provide adequate accuracy for clinical applications.