Hybrid machine learning to localize atrial flutter substrates using the surface 12-lead electrocardiogram

Abstract

Aims: Atrial flutter (AFlut) is a common re-entrant atrial tachycardia driven by self-sustainable mechanisms that cause excitations to propagate along pathways different from sinus rhythm. Intra-cardiac electrophysiological mapping and catheter ablation are often performed without detailed prior knowledge of the mechanism perpetuating AFlut, likely prolonging the procedure time of these invasive interventions. We sought to discriminate the AFlut location [cavotricuspid isthmus-dependent (CTI), peri-mitral, and other left atrium (LA) AFlut classes] with a machine learning-based algorithm using only the non-invasive signals from the 12-lead electrocardiogram (ECG).

Methods and results: Hybrid 12-lead ECG dataset of 1769 signals was used (1424 in silico ECGs, and 345 clinical ECGs from 115 patients-three different ECG segments over time were extracted from each patient corresponding to single AFlut cycles). Seventy-seven features were extracted. A decision tree classifier with a hold-out classification approach was trained, validated, and tested on the dataset randomly split after selecting the most informative features. The clinical test set comprised 38 patients (114 clinical ECGs). The classifier yielded 76.3% accuracy on the clinical test set with a sensitivity of 89.7%, 75.0%, and 64.1% and a positive predictive value of 71.4%, 75.0%, and 86.2% for CTI, peri-mitral, and other LA class, respectively. Considering majority vote of the three segments taken from each patient, the CTI class was correctly classified at 92%.

Conclusion: Our results show that a machine learning classifier relying only on non-invasive signals can potentially identify the location of AFlut mechanisms. This method could aid in planning and tailoring patient-specific AFlut treatments.

Keywords: Atrial flutter; Cardiac modelling; Electrocardiography; Machine learning; Personalized medicine.

Graphical Abstract

Figure 1

Example of clinical CTI-dependent, peri-mitral, and other LA AFlut 12-lead ECGs in the test set, respectively. The red segments represent one of the three AFlut single cycles extracted and used in this work for this specific patient. The values of the two most predictive metrics are reported for the three cases in example ($\mathit{FwD}$ and $E_{{\mathit{icRQA}}_{PC3}}^{VL}$).

Figure 2

Example of different re-entry paths for each class in simulation with the respective 12-lead ECG. The top line shows a frame of one of the cardiac simulations computed in this study: CTI-dependent flutter with counterclockwise direction (left), peri-mitral flutter wit clockwise direction (middle), other left atrium flutter, i.e. figure-8 macro-re-entry with anterior direction (right). Following, in the bottom line there are three 12-lead ECGs as result of the top simulations. It can be seen that there is no QRS-T activity given the lack of ventricles from the simulations. The segments used for each signal have been highlighted in red. The values of the two most predictive metrics are reported for the three cases in example ($\mathit{FwD}$ and $E_{{\mathit{icRQA}}_{PC3}}^{VL}$).

Figure 3

Clinical CTI-dependent case with atypical ECG-features (positive atrial waves in lead II, III, aVF, and V1 to V6) correctly classified by the classifier. The red segments represent one of the three AFlut single cycles extracted and used in this work for this specific patient. The values of the two most predictive metrics are reported ($\mathit{FwD}$ and $E_{{\mathit{icRQA}}_{PC3}}^{VL}$).