ICA-Based EEG denoising: a comparative analysis of fifteen methods

Abstract

Abstract. Independent Component Analysis (ICA) plays an important role in biomedical engineering. Indeed, the complexity of processes involved in biomedicine and the lack of reference signals make this blind approach a powerful tool to extract sources of interest. However, in practice, only few ICA algorithms such as SOBI, (extended) InfoMax and FastICA are used nowadays to process biomedical signals. In this paper we raise the question whether other ICA methods could be better suited in terms of performance and computational complexity. We focus on ElectroEncephaloGraphy (EEG) data denoising, and more particularly on removal of muscle artifacts from interictal epileptiform activity. Assumptions required by ICA are discussed in such a context. Then fifteen ICA algorithms, namely JADE, CoM2, SOBI, SOBI rob , (extended) InfoMax, PICA, two different implementations of FastICA, ERICA, SIMBEC, FOBIUMJAD, TFBSS, ICAR3, FOOBI1 and 4-CANDHAPc are briefly described. Next they are studied in terms of performance and numerical complexity. Quantitative results are obtained on simulated epileptic data generated with a physiologically-plausible model. These results are also illustrated on real epileptic recordings.

EEG extended source localization: Tensor-based vs. conventional methods

“…Assuming that artifacts have already been removed from the data as can be achieved by methods as described, for instance, in (Comon and Jutten, 2010;Albera et al, 2012), all dipoles that do not belong to an extended source can be considered to emit background activity. Consequently, in order to distinguish between the extended sources, that we are looking for, and the noisy background activity, we can rewrite the data model (1) in the following way:…”

An Efficient Jacobi-Like Deflationary ICA Algorithm: Application to EEG Denoising

“…Note that, unlike the three other algorithms, RobustICA does not require any prewhitening. The performance was computed as a function of computational complexity using the Normalized Mean-Squared Error (NMSE) as defined in [1]. In our experiment, the data length is fixed to 5120 samples and the SNR value is −5 dB.…”

“…(1) (t) as a product of P − 1 elementary Givens rotations G (1) p (t p ) and of sequentially identifying the P − 1 corresponding parameters t p . The (P − 1)-dimensional optimization problem is thus replaced with P −1 sequential mono-dimensional optimization problems.…”

“…The N 4 FO cumulants C n1,n2,n3,n4,x of vector x can be estimated at the beginning of the procedure and sorted in a (N 2 × N 2 ) matrix, Q x , called quadricovariance [1]. The FO cumulants C n1,n2,n3,n4,x (p) of vector x (p) can then be derived using the following formula Q x (p) = HQ x H T where:…”

“…NDEPENDENT Component Analysis (ICA) [8], [9] is a very useful tool in signal processing especially to process biomedical signals such as ElectroEncephaloGraphic (EEG) data [1]- [5]. The ICA problem consists of retrieving unobserved realizations of a P -dimensional random vector s = [s 1 , .…”