Machine learning-enhanced MCG for LVH detection: a multi-domain feature selection approach

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Left ventricular hypertrophy (LVH) (n=481). Intervention: XGBoost-Focal Loss machine learning framework using magnetocardiography (MCG) signals vs. Conventional machine learning baselines (e.g., Logistic Regression). Primary outcome: Area Under the Curve (AUC) for LVH detection on independent validation set.

Q: What does this research mean for the field?

An interpretable machine learning framework analyzing multi-domain magnetocardiography (MCG) signals accurately detects left ventricular hypertrophy (LVH), identifying T-wave magnetic polarity as the most influential predictor. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Xiaoxia Chen; Xuanhao Xu; Hong Shen; Jinyang Wang; Y. F. Wang; Lei Xia; Jiaqi Wang; Meifang Gao; C X Shen; Yukun Luo

doi:10.1007/s00371-026-04433-x

March 31, 2026The Visual ComputerOpen Access

Machine learning-enhanced MCG for LVH detection: a multi-domain feature selection approach

View Full Paper

Ask AI

Bookmark

Share

Key result

An interpretable machine learning framework using XGBoost with Focal Loss achieved an AUC of 0.837 for detecting left ventricular hypertrophy from magnetocardiography signals in an independent validation set.

Why the study?

Magnetocardiography provides high-resolution spatiotemporal insights into cardiac electrophysiology but remains underutilized for LVH diagnosis due to a lack of interpretable analytical tools.

Does a machine learning-enhanced magnetocardiography (MCG) framework improve the diagnostic accuracy for detecting left ventricular hypertrophy?

Key Points

Develop and validate an interpretable machine learning framework to identify left ventricular hypertrophy from multi-domain magnetocardiography signals.
Analyzed magnetocardiography data from 481 subjects across four domains: temporal waves, spatial waves, current source imaging, and dynamic characterization.
Trained an XGBoost model optimized with Focal Loss to address class imbalance, applying multi-stage feature selection including the Boruta algorithm.
Evaluated performance using fivefold cross-validation, an independent holdout validation set comprising the most recently enrolled 30% of subjects, and SHAP interpretability analysis.
The XGBoost-Focal Loss model achieved an AUC of 0.902 ± 0.036, an accuracy of 0.843 ± 0.032, a sensitivity of 0.599 ± 0.116, and a specificity of 0.942 ± 0.028 on fivefold cross-validation.
In independent validation, the model attained an AUC of 0.837 with 89.89% specificity and 74.55% sensitivity, outperforming logistic regression (AUC 0.735), random forest (AUC 0.775), and gradient boosting (AUC 0.762).
SHAP analysis identified T-wave magnetic polarity features as the most influential predictors, reflecting electrical repolarization alterations in hypertrophied heart tissue.

Study Design

Type

Observational (n=481)

Multicenter

No

Structured PICO

Does a machine learning-enhanced magnetocardiography (MCG) framework improve the diagnostic accuracy for detecting left ventricular hypertrophy?

P

Population

481 subjects evaluated for left ventricular hypertrophy (LVH)

I

Intervention

Machine learning-enhanced Magnetocardiography (MCG) using an XGBoost-Focal Loss framework

C

Comparator

Conventional machine learning baselines (Logistic Regression, Random Forest, Extra Trees, Gradient Boosting)

O

Outcome

Diagnostic accuracy for left ventricular hypertrophy (LVH) detection, measured by Area Under the Curve (AUC)surrogate

Main Result

Absolute Event Rate: 0.837% vs 0.735%

An interpretable machine learning framework using magnetocardiography (MCG) provides high diagnostic accuracy for automated left ventricular hypertrophy detection, identifying T-wave magnetic polarity as a key predictor.

Limitations

Relatively limited sample size might impact model generalization and the ability to identify rare LVH subtypes
A performance drop was observed on the validation set (AUC decreased from 0.902 to 0.837)
Lack of prospective, multi-center studies to further calibrate the model for diverse clinical environments
Relatively limited sample size
Drop in performance on the validation set (AUC decreased from 0.902 to 0.837)
Lack of prospective, multi-center validation
Need to map MCG model-identified features to specific pathophysiological mechanisms

Implication

May improve automated LVH detection with MCG in specialized clinics; extends prior MCG research with interpretable ML frameworks.

Abstract

Magnetocardiography (MCG) provides high-resolution spatiotemporal insights into cardiac electrophysiology but remains underutilized for left ventricular hypertrophy (LVH) diagnosis due to a lack of interpretable analytical tools. We propose a novel interpretable machine learning framework that systematically decodes MCG signals across four complementary domains: temporal waves, spatial waves, current source imaging, and dynamic characterization. To address class imbalance, we integrated Focal Loss into the XGBoost objective function. In a dataset of 481 subjects, our model achieved an AUC of 0.902 in cross-validation and 0.837 in independent validation, significantly outperforming conventional baselines. Notably, SHapley Additive exPlanations (SHAP) identified T-wave magnetic polarity as the most influential predictor, offering new perspectives on the electrophysiological remodeling of hypertrophied myocardium. This framework bridges the gap between raw sensor data and clinical decision-making, providing a robust tool for automated LVH detection.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Chen et al. (2026) conducted an observational in Left ventricular hypertrophy (LVH) (n=481). XGBoost-Focal Loss machine learning framework using magnetocardiography (MCG) signals vs. Conventional machine learning baselines (e.g., Logistic Regression) was evaluated on Area Under the Curve (AUC) for LVH detection on independent validation set. An interpretable machine learning framework using XGBoost with Focal Loss achieved an AUC of 0.837 for detecting left ventricular hypertrophy from magnetocardiography signals in an independent validation set.

synapsesocial.com/papers/69cf5e115a333a821460c39b https://doi.org/10.1007/s00371-026-04433-x

View Full Paper

Ask AI

Bookmark

Share

Machine learning-enhanced MCG for LVH detection: a multi-domain feature selection approach

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Machine learning-enhanced MCG for LVH detection: a multi-domain feature selection approach

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading