P3655Predictive accuracy of cardiac magnetic field dynamics and inverse solution to non-invasively detect ischemic heart disease with unshielded magnetocardiographic mapping at rest

Abstract

Background: Magnetocardiographic mapping (MCG) is increasingly reported as a sensitive, radiation-free, method for non-invasive diagnosis of ischemic heart disease (IHD). Over 1000 IHD patients have been investigated in unshielded hospital laboratories (mostly with cryogenic instrumentation, and recently also with a portable multi-induction-coil sensor array). Reported predictive accuracy (PA) is variable as a function of study design, patients (pts) selection (chronic IHD versus acute chest pain) and analytic methods. This retrospective study aimed to assess the PA of MCG in detecting IHD, using automatic analysis of T-wave magnetic field (MF) parameters. Method(s): Rest MCG data of 136 pts (mean age: 62+/-10 years, 32.5% females) with suspected IHD (all chest-pain free during MCG scan) were analyzed and compared with those of 72 healthy subjects (HS) as control group. 10 patients with previous NSTEMI, but no evidence of ischemia at nuclear scan when MCG was performed, were excluded. Unshielded MCG was obtained with a 36-channel SQUID system (intrinsic sensitivity 30 fT/ Hz, above 1Hz). The MCG recording time was typically 300 seconds. IHD was confirmed (IHD+) or excluded (IHD-), with effort ECG, stress-SPECT, and/or with coronary angiography, as appropriate. Thirteen MCG parameters were derived from automatic analysis of the dynamics of T-wave MF extrema (5 parameters) and of the spatial dynamics of the effective magnetic vector (EMV) (8 parameters), automatically calculated after inverse solution with the equivalent magnetic dipole model (EMDis). Two different comparisons were performed: - the first, between IHD+ pts and a pooled group of HS and IHD- pts - the second, between IHD+ and IHD- pts Analysis of variance and regularized logistic regression modeling were employed to study the relationship between the response variable (pts vs controls) and the 13 MCG parameters. Result(s): According to clinical workup, 60 pts were IHD+ and 66 IHD-. Assuming an IHD prevalence of 12%, MCG overall sensitivity, specificity, positive (PPV) and negative (NPV) predictive values were 72%, 93%, 60% and 96%, respectively, in differentiating IHD+ pts from IHD- and HS. PA was 87%. These results were also confirmed when differentiating IHD+ from IHD- (sensitivity 75%, sensibility 86%, PPV 43%, NPV 96%). Exclusion of pts with LBBB did not affect MCG PA in IHD+ pts. Conclusion(s): Automatic analysis of the dynamics of T-wave MF and of the EMV parameters, obtained from rest unshielded MCG, provided an impressive NPV over 95%. Thus, MCG may be useful for early triage of pts with acute chest pain and still non-diagnostic ECG and enzyme patterns, to rule-out IHD- pts. A better PA is foreseen by combining in a machine learning approach also parameters derived from the analysis of QRS and ST MF parameters.