Session M1.4

Effect of Respiration on the Solutions of Forward and Inverse Electrocardiographic Problems: A Simulation Study

Y Jiang*, Y Meng, D Farina, O Dössel

Karlsruhe Institute of Technology
Karlsruhe, Germany

The forward problem of electrocardiography aims at obtaining a better understanding of cardiac electrophysiological activities, by means of computer modeling and simulation. Whereas, the inverse electrocardiographic problem provides a direct insight of electrical sources into the heart without interventional procedures. Nowadays, forward and inverse problems of electrocardiography are mostly investigated in static thorax models, which are usually built from MRI or CT data-sets with the heart in diastolic state and the lungs in deflated state. Because heart motion and respiration are not taken into account, modeling errors are introduced into the system, which can cause inaccuracy in forward simulation and instability in inverse solution. Besides heart motion, neglecting respiration may also lead to remarkable uncertainties in both forward and inverse solutions.
In the present investigation a dynamic lung model considering change in volume and conductivity of lungs is developed to study the influence of respiration on simulated ECG and inverse reconstructions. A respiratory cycle of 4s including 4 normal cardiac cycles with a duration of 1s each is simulated. The simulation is performed in both static and dynamic models. From the simulated BSPMs and standard ECG leads, noticeable difference can only be seen in the period, in which the lungs approach the maximum in volume and the minimum in conductivity. The variations caused by respiration are mainly in the signal amplitude, whereas the patterns of BSPM and ECG do not show significant change. Nevertheless, respiration should be taken into account when high accuracy in ECG simulation is desired.
Tikhonov 0-order regularization and GMRES method are applied to solve the inverse problem for epicardial potentials in both dynamic and static models. Respiration has non-negligible effect on inverse solutions. The reconstructions using Tikhonov 0-order without modeling error (introduced by neglecting respiration) are considerably better than those obtained with modeling error and an improvement of up to 0.15 in correlation coefficient is shown. Although Krylov subspace methods like the GMRES method are relatively stable against modeling errors, the inverse solutions can still be improved if the modeling error from disregarding respiration is eliminated. Therefore, it is suggested to incorporate the respiration into the inverse calculation according to the results in the present study.

(Abstract Control Number: 110)