A Bayesian method to estimate single-trial event-related potentials with application to the study of the P300 variability

• Published in: Journal of neuroscience methods 2011-05, Vol.198 (1), p.114-124
• Main Authors: D’Avanzo, Costanza, Schiff, Sami, Amodio, Piero, Sparacino, Giovanni
• Format: Article
• Language: English
• Subjects: Adult; Bayes estimation; Bayes Theorem; Brain - physiology; Computer Simulation; Electroencephalography - methods; Event-Related Potentials, P300 - physiology; Evoked response; Female; Humans; Male; Models, Biological; Optimal filtering; P300; Reaction Time; Single-trial; Young Adult
• ISSN: 0165-0270; 1872-678X
• DOI: 10.1016/j.jneumeth.2011.03.010

► A two-stage Bayesian approach for the single-trial estimation of ERPs is proposed. ► Simulated studies confirm its capability in tracking ERPs variability. ► Satisfactory results in terms of MSE have been obtained also at low SNRs. ► Good performance in the determination of the P300 parameters, latency in particular. ► Results on real data encourage the application of the technique. We propose a Bayesian method to extract single-trial event related potentials (ERPs). The method is formulated in two stages. In the first stage, each of the N raw sweeps is processed by an individual “optimal” filter, where the 2nd order a priori statistical information on the background EEG and on the unknown ERP is, respectively, estimated from pre-stimulus data and obtained through the multiple integration of a white noise process model which is identifiable from post-stimulus data thanks to a smoothing criterion. Then, a mean ERP is determined as the weighted average of the filtered sweeps, where each weight is inversely proportional to the expected value of the norm of the correspondent filter error. In the second stage, single-sweep estimation is dealt with within the same framework, by using the average ERP estimated in the previous stage as a priori expected response. The method is successfully tested on simulated data and then employed on real data with the aim of investigating the variability of the P300 component during a cognitive visual task. A comparison with other literature methods is also performed. Results encourage further use of the proposed method to investigate if and how diseases, e.g., cirrhosis, are associated to differences in P300 variability.