Expert-enhanced machine learning for cardiac arrhythmia classification.

We propose a new method for the classification task of distinguishing atrial fibrillation (AFib) from regular atrial tachycardias including atrial flutter (AFlu) based on a surface electrocardiogram (ECG). Recently, many approaches for an automatic classification of cardiac arrhythmia were proposed and to our knowledge none of them can distinguish between these two. We discuss reasons why deep learning may not yield satisfactory results for this task. We generate new and clinically interpretable features using mathematical optimization for subsequent use within a machine learning (ML) model. These features are generated from the same input data by solving an additional regression problem with complicated combinatorial substructures. The resultant can be seen as a novel machine learning model that incorporates expert knowledge on the pathophysiology of atrial flutter. Our approach achieves an unprecedented accuracy of 82.84% and an area under the receiver operating characteristic (ROC) curve of 0.9, which classifies as "excellent" according to the classification indicator of diagnostic tests. One additional advantage of our approach is the inherent interpretability of the classification results. Our features give insight into a possibly occurring multilevel atrioventricular blocking mechanism, which may improve treatment decisions beyond the classification itself. Our research ideally complements existing textbook cardiac arrhythmia classification methods, which cannot provide a classification for the important case of AFib↔AFlu. The main contribution is the successful use of a novel mathematical model for multilevel atrioventricular block and optimization-driven inverse simulation to enhance machine learning for classification of the arguably most difficult cases in cardiac arrhythmia. A tailored Branch-and-Bound algorithm was implemented for the domain knowledge part, while standard algorithms such as Adam could be used for training.

Discriminating atrial flutter from atrial fibrillation using a multilevel model of atrioventricular conduction.

BACKGROUND: The discrimination between atrial flutter (AFlu) and atrial fibrillation (AFib) can be made difficult by an irregular ventricular response owing to complex conduction phenomena within the atrioventricular (AV) node, known as multilevel AV block. We tested the hypothesis that a mathematical algorithm might be suitable to discriminate both arrhythmias. OBJECTIVES: To discriminate AFlu with irregular ventricular response from AFib based on the sequence of R-R intervals. METHODS: Intracardiac recordings of 100 patients (50 patients with AFib and 50 patients with AFlu) were analyzed. On the basis of a numerical simulation of variable flutter frequencies followed by 2 levels of AV block in series, a given sequence of R-R intervals was analyzed. RESULTS: Although the ventricular response displays absolute irregularity in AFib, the sequences of R-R intervals follow certain rules in AFlu. We find that using a mathematical simulation of multilevel AV block, based on the R-R sequence of 16 ventricular beats, a stability of atrial activation could be predicted with a sensitivity of 84% and a specificity of 74%. When limiting the ventricular rate to 125 beats/min, discrimination could be performed with a sensitivity of even 89% and a specificity of 80%. In cases of AFlu, the atrial cycle length could be predicted with high accuracy. CONCLUSION: On the basis of the electrophysiological mechanism of multilevel AV block, we developed a computer algorithm to discriminate between AFlu and Afib. This algorithm is able to predict the stability and cycle length of atrial activation for short R-R sequences with high accuracy.