MCG-Net: End-to-end Fine-grained Delineation and Diagnostic Classification of Cardiac Events from Magnetocardiographs

In this paper, we propose an end-to-end deep learning architecture, referred as MCG-Net, integrating convolutional neural network (CNN) with transformer-based global context block for fine-grained delineation and diagnostic classification of four cardiac events from magnetocardiogram (MCG) data, namely Q-, R-, S- and T-waves. MCG-Net} takes advantage of a multi-resolution CNN backbone as well as the state-of-the-art (SOTA) transformer encoders that facilitate global temporal feature aggregation. Besides the novel network architecture, we introduce a multi-task learning scheme to achieve simultaneous delineation and classification. Specifically, the problem of MCG delineation is formulated as multi-class heatmap regression. Meanwhile, a binary diagnostic classification label as well as a duration are jointly estimated for each cardiac event using features that are temporally aligned by event heatmaps. The framework is evaluated on a clinical MCG dataset, containing data collected from 270 subjects with cardiac anomalies and 108 control subjects. We designed and conducted a two-fold cross-validation study to validate the proposed method and to compare its performance with the SOTA methods. Experimental results demonstrated that our method outperformed counterparts on both event delineation and diagnostic classification tasks, achieving respectively an average ECG-F1 of 0.987 and an average Event-F1 of 0.975 for MCG delineation, and an average accuracy of 0.870, an average sensitivity of 0.732, an average specificity of 0.914 and an average AUC of 0.903 for diagnostic classification. Comprehensive ablation experiments are additionally performed to investigate effectiveness of different network components.

An Integrated Maximum Current Density Approach for Noninvasive Detection of Myocardial Infarction

We present a new approach of integrated maximum current density (IMCD) for the noninvasive detection of myocardial infarction (MI) using magnetocardiography (MCG) data acquired from a superconducting quantum interference device (SQUID) system. In this paper, we investigated the relationship of the maximum current density (MCD) in the current density map and the underlying equivalent current dipole (ECD) based on a novel method of reconstructing the ECD in the extremum circle of the magnetic field map. The performance of IMCD and the integrated ECD (IECD) approaches were also evaluated by using 61-channel MCG data from 39 healthy subjects and 102 patients with ST elevation myocardial infarction (STEMI). Statistical analysis of the healthy and STEMI groups demonstrate that the IMCD approach obtains sensitivity and specificity up to 91.2% and 84.6%, somewhat higher than that of IECD, respectively. The results indicate that IMCD provides spatiotemporal information regarding cardiac electrical activity during ventricular repolarization. This approach may be helpful to diagnose MI in clinic application. The physical concept of the approach is also explained in this paper.

Early detection of coronary artery disease in patients studied with magnetocardiography: an automatic classification system based on signal entropy

We propose an automatic system for the classification of coronary artery disease (CAD) based on entropy measures of MCG recordings. Ten patients with coronary artery narrowing ≥ or ≤ 50% were categorized by a multilayer perceptron (MLP) neural network based on Linear Discriminant Analysis (LDA). Best results were obtained with MCG at rest: 99% sensitivity, 97% specificity, 98% accuracy, 96% and 99% positive and negative predictive values for single heartbeats. At patient level, these results correspond to a correct classification of all patients. The classifier's suitability to detect CAD-induced changes on the MCG at rest was validated with surrogate data.

Changes in dipolar structure of cardiac magnetic field maps after ST elevation myocardial infarction

BACKGROUND

Pathological changes in cardiac electrophysiology have been investigated in coronary artery disease using magnetocardiography. Aim of this work was to examine the structure of cardiac magnetic field maps (MFM) during ventricular depolarization and repolarization in patients with acute ST elevation myocardial infarction (STEMI).

METHODS

Magnetocardiograms were recorded in 39 healthy subjects and 97 patients who had been successfully revascularized after STEMI. Using the Karhunen-Loeve transform, 12 eigenmaps were constructed for six intervals within the QT interval of each subject's signal-averaged data. The relative information content of the eigenmaps was compared between STEMI patients and healthy subjects.

RESULTS

Relative nondipolar content was between 0.03% and 0.52% higher in the STEMI group, (P < 0.001 for the repolarization intervals). Information content of the first dipolar eigenmap in the STEMI group was reduced by 2.6%-11.7% (P < 0.001 for the repolarization intervals). STT interval was best able to discriminate between groups: area-under-the-curve for nondipolar content was 85.8% (P < 0.001), for the first eigenmap 91.7% (P < 0.001). Severity of infarction was reflected in lower STT interval map 1 content for patients with anterior versus posterior infarction (83%± 11% vs. 87%± 10%, P < 0.05), with wall motion disturbances (84%± 11% vs. 92%± 7%, P < 0.001) and with microvascular obstruction (81%± 12% vs. 87%± 10%, P < 0.05). Regression analysis showed that patients with lower ejection fraction tended to have less information content (P < 0.001).

CONCLUSION

STEMI is associated with a loss of spatial coherence during repolarization, as quantified by principal component analysis of cardiac MFM.