Cardiac magnetic field mapping quantified by Kullback-Leibler entropy detects patients with coronary artery disease

Cardiac magnetic field mapping (CMFM) is a noninvasive method to determine cardiac electrical activity. We analysed the utility of CMFM for the detection of patients with coronary artery disease (CAD) without subjecting them to stress. We studied 59 healthy control subjects and 101 patients with CAD without previous myocardial infarction (MI). The heart's magnetic field was recorded over the anterior chest wall using a multichannel magnetic measurement system with axial second-order gradiometers. The evaluation of CMFM was based on comparison of the 'ideal' group mean maps of young healthy subjects and maps of examined individuals. Three measures of similarity were considered: Kullback-Leibler (KL) entropy, normalized residual magnetic field strength and deviations in the magnetic field map orientation. The mean values of these parameters during the depolarization and repolarization were used for further classification with the help of logistic regression. The feature set based on the KL-entropy demonstrated the best classification results (sensitivity/specificity of 85/80%), followed by the residual feature (85/75%) and the magnetic field orientation feature (80/73%) sets. The forward stepwise technique was applied to select the best set of features from the combined feature set. Two parameters were selected, namely the KL-entropy for the repolarization period and the residual parameter for the depolarization period. The classification based on these parameters demonstrated a sensitivity of 88% and a specificity of 88% for the distinction of CAD patients from the control subjects. The area under the receiver operator curve was 94%. Hence, we suggest that CMFM evaluation based on KL-entropy is a promising technique to identify patients with CAD.