Cardiomagnetic source imaging

Abstract

Magnetocardiographic (MCG) source imaging has received increasing interest in recent years. With a high enough localization accuracy of the current sources in the heart, valuable information can be provided, e.g., for the pre-ablative evaluation of arrhythmia patients. Furthermore, preliminary studies indicate that ischemic areas, i.e. areas which are suffering from lack of oxygen, and infarcted regions could be localized from multichannel MCG recordings. In this thesis, the accuracy of cardiomagnetic source imaging results, obtained by using different current source models, was investigated. In addition, the effect of the torso model on the localization accuracy was examined. In some studies, also body surface potential mapping (BSPM) data were used for comparison purposes. A high impact was given to clinical validation, i.e. how the calculation methods would work in patients.

The equivalent current dipole (ECD) source model was found to produce accurate (within 3-11 mm) localizations of focal current sources in a thorax phantom and in 15 patients with a non-magnetic stimulation catheter in the heart. The accuracy was found to depend on the signal-to-noise ratio and on the goodness of fit of the localizations. The corresponding accuracy determined from simultaneous multichannel BSPM recordings in 10 patients was 25 mm. In order to localize wider source regions in the heart, distributed source models were also investigated in the thesis. Current density estimates (CDEs) were calculated in the catheter patients and in 13 patients with coronary artery disease (CAD). Promising results were obtained by using second-order Tikhonov regularization in the calculations. CDEs were found to localize both myocardial ischemia in single-vessel CAD patients, as well as more complex chronic ischemia in three-vessel CAD patients.

In addition to the ECD and CDE source models, the uniform double layer (UDL) model was used in the source imaging studies. With the UDL model, the whole depolarization of the ventricles can be represented with a single inverse solution. In the validation of the activation time maps calculated from MCG and BSPM recordings, invasively measured epicardial electrograms were used to construct the reference epicardial activation times. The overall patterns of activation in the reference data were reproduced relatively well in the calculated activation time maps.

The high quality of the inverse solutions obtained in this thesis prompts the use of cardiomagnetic source imaging in several clinical applications, such as in electrophysiological studies and in the estimation of myocardial viability.