Ischemic heart disease recognition by k-NN classification of current density distribution maps

Magnetocardiography is non-invasive and risk-free technique allowing body-surface recording of the magnetic fields generated by the electrical activity of the heart. In this paper, k-Nearest Neighbor algorithm is applied for binary classification of myocardium current density distribution maps (CDDM). Different types of CDDMs from patients with ischemic heart disease are compared with normal subjects. Selection of number of neighbors for k-NN classifier was performed to optimize classification characteristics. Specificity, accuracy, precision and sensitivity of classification as functions of number of neighbors in k-NN are obtained. Depending on CDDMs type, accuracy in a range of 60-90%, 32-88% sensitivity, 78-95% specificity and 77-93% precision were achieved. Considering all studied groups of patients with ischemic heart disease, optimal number of neighbors for obtaining highest accuracy using for k-NN classification lies in a range of 20-25 neighbors.