M. Milanesi

Independent component analysis applied to the removal of motion artifacts from electrocardiographic signals

Electrocardiographic (ECG) signals are affected by several kinds of artifacts that may hide vital signs of interest. In this study we apply independent component analysis (ICA) to isolate motion artifacts. Standard or instantaneous ICA, which is currently the most addressed ICA model within the context of artifact removal, is compared to two other ICA techniques. The first technique is a frequency domain approach to convolutive mixture separation. The second is based on temporally constrained ICA, which enables the estimation of only one component close to a particular reference signal. Performance indexes evaluate ECG complex enhancement and relevant heart rate errors. Our results show that both convolutive and constrained ICA implementations perform better than standard ICA, thus opening up a new field of application for these two methods. Moreover, statistical analysis reveals that constrained ICA and convolutive ICA do not significantly differ concerning heart rate estimation, even though the latter overcomes the former in ECG morphology recovery.

Multichannel techniques for motion artifacts removal from electrocardiographic signals.

Electrocardiographic (ECG) signals are affected by several kinds of artifacts, that may hide vital signs of interest. Motion artifacts, due to the motion of the electrodes in relation to patient skin, are particularly frequent in bioelectrical signals acquired by wearable systems. In this paper we propose different approaches in order to get rid of motion confounds. The first approach we follow starts from measuring electrode motion provided by an accelerometer placed on the electrode and use this measurement in an adaptive filtering system to remove the noise present in the ECG. The second approach is based on independent component analysis methods applied to multichannel ECG recordings; we propose to use both instantaneous model and a frequency domain implementation of the convolutive model that accounts for different paths of the source signals to the electrodes

Frequency domain approach to blind source separation in ECG monitoring by wearable system

In this paper we present a method for removing artifacts from biomedical signals acquired by wearable systems, taking advantage of multichannel data acquisition since both artifacts and signals of interest show common features in different channels. In order to take into account the effects of the different paths from the source signals to the sensors, we propose a method based on blind separation of convolutive mixtures: the observed data are seen as linear mixtures of filtered source signals where neither the source signals nor the convolution and mixing processes are known. The only hypothesis we make to recover the original sources is the statistical independence among them. The proposed method was applied on real ECG signals corrupted by motion artifacts with satisfactory results

Residual Dependency Estimation of Independent Components Applied to EEG Event Related Potentials Associated with Emotional Processing

Independent component analysis can be employed as an exploratory method in electroencephalographic (EEG) data analysis. However, the assumption of statistical independence among the estimated components is not always fulfilled by ICA-based numerical methods. Furthermore it may happen that one physiological source can be split in two or more components. As a consequence, the estimated components must be further investigated to assess the existence of reciprocal similarities. In this work a method for finding residual dependency subsets of component is proposed. Firstly a hierarchical clustering stage is carried out to classify ICA results. Then the hierarchical tree is investigated at each level by two indices to evaluate the tightness of all clusters. At the same time clustered scalp projections are compared with a template, which is shaped by applying ensemble ICA to a training dataset. Results are shown on EEG data acquired in event-related brain potentials (ERPs) studies for emotional pictures processing. In this kind of experiment ERPs are measured whilst unpleasant and neutral images are shown to a subject. The clustering procedure and the performance indices succeeded in isolating compact groups of components. These components, taken together, reflect the brains biopotentials related to emotional processing at different cortical areas.