Magnetocardiography-Based Bundle Branch Block Detection Using Machine Learning Methods

The potential of magnetocardiography (MCG) in diagnosing bundle branch block(BBB) has been preliminarily validated, but studies on diagnostic accuracy are lacking. Using electrocardiogram (ECG) diagnoses observed by senior medical experts as the reference standard, we constructed a machine learning model incorporating seven MCG features. We evaluated the performance of two machine learning models using 5-fold cross-validation and a validation set. In a cohort of 110 healthy controls and 116 patients with BBB, both the Random Forest model (AUC = 0.987) and the Gradient Boosting model (AUC = 0.991) demonstrated excellent diagnostic accuracy. The machine learning method proposed in this study, based on MCG, provides clinicians with a rapid and accurate tool for diagnosing BBB, potentially enhancing the acceptance of MCG in clinical diagnosis. Future work will explore the integration of additional features based on magnetic pole direction and the adoption of deep learning methods—two strategies expected to yield significant gains in the model’s classification accuracy and clinical applicability.