Comparison of BSS algorithms in SMCG data

The field distribution generated by the heart and detected by magnetocardiography (MCG) shows strong effects, comparing the field distribution at rest and under stress. To study the possible clinical application, it is convenient to avoid pharmacological stress and perform Stress MCG (SMCG) using conventional physical stress with an ergometer. Unfortunately the subject movements and the ergometer residual artefacts produce extra noise and disturbances which should be eliminated. For this purpose 4 different Blind Source Separation (BSS) algorithms were tested: fastICA, TDSEP, JADE and SHIBBS. The algorithms were applied on 12 SMCG data sets for the task to de-noise them. The correlation between odd and even beats (using the simultaneously recorded electrocardiogram (ECG)) and the known spectral content were considered for automatically detecting noise and disturbances. All the methods were able to detect the ergometer noise. The best results were achieved by TDSEP, both in computation time and in goodness of separation. This work shows that, with BSS, the extraction of heart signals from ergometer SMCG data is becoming feasible.