Clinical validation of machine learning for automatic analysis of multichannel magnetocardiography

Authors:
Riccardo Fenici;Donatella Brisinda;Anna Maria Meloni;Karsten Sternickel;Peter Fenici
Affiliations:
Clinical Physiology – Biomagnetism Research Center, Catholic University of Sacred Heart, Rome, Italy;Clinical Physiology – Biomagnetism Research Center, Catholic University of Sacred Heart, Rome, Italy;Clinical Physiology – Biomagnetism Research Center, Catholic University of Sacred Heart, Rome, Italy;Clinical Physiology – Biomagnetism Research Center, Catholic University of Sacred Heart, Rome, Italy;Clinical Physiology – Biomagnetism Research Center, Catholic University of Sacred Heart, Rome, Italy
Venue:
FIMH'05 Proceedings of the Third international conference on Functional Imaging and Modeling of the Heart
Year:
2005

Citing 2
Cited 1

Kernel partial least squares regression in reproducing kernel hilbert space

The Journal of Machine Learning Research
First 36-channel magnetocardiographic study of CAD patients in an unshielded laboratory for interventional and intensive cardiac care

FIMH'03 Proceedings of the 2nd international conference on Functional imaging and modeling of the heart

Identification of ischemic heart disease via machine learning analysis on magnetocardiograms

Computers in Biology and Medicine

Quantified Score

Hi-index	0.00

Visualization

Abstract

Magnetocardiographic (MCG) mapping measures magnetic fields generated by the electrophysiological activity of the heart. Quantitative analysis of MCG ventricular repolarization (VR) parameters may be useful to detect myocardial ischemia in patients with apparently normal ECG. However, manual calculation of MCG VR is time consuming and can be dependent on the examiner’s experience. Alternatively, the use of machine learning (ML) has been proposed recently to automate the interpretation of MCG recordings and to minimize human interference with the analysis. The aim of this study was to validate the predictive value of ML techniques in comparison with interactive, computer-aided, MCG analysis. ML testing was done on a set of 140 randomly analysed MCG recordings from 74 subjects: 41 patients with ischemic heart disease (IHD) (group 1), 32 of them untreated (group 2), and 33 subjects without any evidence of cardiac disease (group 3). For each case at least 2 MCG datasets, recorded in different sessions, were analysed. Two ML techniques combined identified abnormal VR in 25 IHD patients (group 1) and excluded VR abnormalities in 28 controls (group 3) providing 75% sensitivity, 85% specificity, 83% positive predictive value, 78% negative predictive value, 80% predictive accuracy This result was for the most part in agreement, but statistically better than that obtained with interactive analysis. This study confirms that ML, applied on MCG recording at rest, has a predictive accuracy of 80% in detecting electrophysiological alterations associated with untreated IHD. Further work is needed to test the ML capability to differentiate VR alterations due to IHD from those due to non-ischemic cardiomyopathies.