| Popis: |
The non-invasive localization of accessory pathway (AP) in patients with Wolff-Parkinson-White (WPW) syndrome is typically performed upon physicians’ diagnoses based on observing their electrocardiogram (ECG) signals, which are not always the same. Therefore, a high-accuracy automatic method can help minimize this gap regarding AP localization. This study was to develop a novel semi-automatic localization of AP in patients with WPW syndrome, using features selected from the cross-recurrence plot (CRP) of consecutive precordial leads on ECG. The study participants comprised of 31 patients with WPW syndrome (aged 8-69, with the mean age of 31.19±14.69, 32.3% female), receiving successful ablation therapy during the first session. The features extracted from the CRP, including laminarity (LAM), trapping time (TT), determinism (DET), and mean length of diagonal line (L) were then analyzed. The feature reduction, The classification and the cross-validation (CV) methods were sequential forward selection (SFS), the k-nearest neighbors (KNN) and the leave-one-out (LOO) respectively. The proposed method could differentiate the right and left APs in the patients with WPW syndrome with the accuracy value of 87% (sensitivity: 93.33%, specificity: 81.25%). These results were achieved by the LAM and L features from the CRP of (V1, V2) and (V3, V4), respectively. |
