Limits...
Multiple linear regression to estimate time-frequency electrophysiological responses in single trials.

Hu L, Zhang ZG, Mouraux A, Iannetti GD - Neuroimage (2015)

Bottom Line: Transient sensory, motor or cognitive event elicit not only phase-locked event-related potentials (ERPs) in the ongoing electroencephalogram (EEG), but also induce non-phase-locked modulations of ongoing EEG oscillations.ERD and ERS reflect changes in the parameters that control oscillations in neuronal networks and, depending on the frequency at which they occur, represent neuronal mechanisms involved in cortical activation, inhibition and binding.This permits within-subject statistical comparisons, correlation with pre-stimulus features, and integration of simultaneously-recorded EEG and fMRI.

View Article: PubMed Central - PubMed

Affiliation: Key Laboratory of Cognition and Personality (Ministry of Education) and Faculty of Psychology, Southwest University, Chongqing, China; Department of Neuroscience, Physiology and Pharmacology, University College London, UK. Electronic address: huli@swu.edu.cn.

No MeSH data available.


Related in: MedlinePlus

Generation of regressors in the TF-MLRd approach.First column: Time-frequency representation of one of the thresholded features (‘ERP’, in this example) of the EEG response, obtained by PCA decomposition with Varimax rotation (see Fig. 1). Second column: TFDs representing plausible variability in latency, frequency, morphology in the time domain, and morphology in the frequency domain were generated by shifting and compressing the thresholded response in an enumerative fashion. Third column: For each source of variability, the plausible responses were re-arranged into vectors, which were subsequently stacked into a data matrix (variability matrix). Fourth column: The eigenvalue plots show the explained variance for each of the first 20 generated PCs, for each variability matrix. Note that the first two PCs explain the largest part of the total variance of each source of variability. Fifth column: Five regressors capturing the average magnitude of the considered TF-feature, and its variability in latency (the second PC in latency-shift variability matrix), in frequency (the second PC in frequency-shift variability matrix), in morphology in the time domain (the second PC in latency-compression variability matrix), and in morphology in the frequency domain (the second PC in frequency-compression variability matrix).
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4401443&req=5

f0020: Generation of regressors in the TF-MLRd approach.First column: Time-frequency representation of one of the thresholded features (‘ERP’, in this example) of the EEG response, obtained by PCA decomposition with Varimax rotation (see Fig. 1). Second column: TFDs representing plausible variability in latency, frequency, morphology in the time domain, and morphology in the frequency domain were generated by shifting and compressing the thresholded response in an enumerative fashion. Third column: For each source of variability, the plausible responses were re-arranged into vectors, which were subsequently stacked into a data matrix (variability matrix). Fourth column: The eigenvalue plots show the explained variance for each of the first 20 generated PCs, for each variability matrix. Note that the first two PCs explain the largest part of the total variance of each source of variability. Fifth column: Five regressors capturing the average magnitude of the considered TF-feature, and its variability in latency (the second PC in latency-shift variability matrix), in frequency (the second PC in frequency-shift variability matrix), in morphology in the time domain (the second PC in latency-compression variability matrix), and in morphology in the frequency domain (the second PC in frequency-compression variability matrix).

Mentions: To obtain an accurate estimate of single trial time-frequency responses, not only their variability in latency and frequency, but also their variability in morphology (both in time and frequency domain) should be taken into account. This has a physiological rationale. For example, in some clinical conditions (e.g., optic neuritis during multiple sclerosis), visual-evoked potentials are “desynchronized”, i.e., their amplitudes are reduced because of increased latency jitter, as well as increased width of single-trial responses (Orssaud, 2003; Pelosi et al., 1997). Latency jitter, as well as trial-by-trial variability in response morphology (i.e., wave width or frequency variability) could thus be important parameters for clinical studies. Therefore, in addition to the basis set of TF-MLR, two more regressors, representing the scaling of the single-trial response in time or frequency domain are considered, which leads to the following TF-MLRd model (Fig. 4):(4)Ftf=k1F1s1t+a1c1f+b1+k2F2s2t+a2c1f+b2+k3F3s3t+a3c1f+b3+εwhere s1, s2 and s3 are the coefficients that determine the compression ratios of the time width of ERP, ERD and ERS of each single-trial TFD compared to those of the average TFD, respectively, while c1, c2 and c3 are the coefficients that determine the compression ratios of the frequency width of ERP, ERD and ERS of each single-trial TFD compared to those of the average TFD, respectively.


Multiple linear regression to estimate time-frequency electrophysiological responses in single trials.

Hu L, Zhang ZG, Mouraux A, Iannetti GD - Neuroimage (2015)

Generation of regressors in the TF-MLRd approach.First column: Time-frequency representation of one of the thresholded features (‘ERP’, in this example) of the EEG response, obtained by PCA decomposition with Varimax rotation (see Fig. 1). Second column: TFDs representing plausible variability in latency, frequency, morphology in the time domain, and morphology in the frequency domain were generated by shifting and compressing the thresholded response in an enumerative fashion. Third column: For each source of variability, the plausible responses were re-arranged into vectors, which were subsequently stacked into a data matrix (variability matrix). Fourth column: The eigenvalue plots show the explained variance for each of the first 20 generated PCs, for each variability matrix. Note that the first two PCs explain the largest part of the total variance of each source of variability. Fifth column: Five regressors capturing the average magnitude of the considered TF-feature, and its variability in latency (the second PC in latency-shift variability matrix), in frequency (the second PC in frequency-shift variability matrix), in morphology in the time domain (the second PC in latency-compression variability matrix), and in morphology in the frequency domain (the second PC in frequency-compression variability matrix).
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4401443&req=5

f0020: Generation of regressors in the TF-MLRd approach.First column: Time-frequency representation of one of the thresholded features (‘ERP’, in this example) of the EEG response, obtained by PCA decomposition with Varimax rotation (see Fig. 1). Second column: TFDs representing plausible variability in latency, frequency, morphology in the time domain, and morphology in the frequency domain were generated by shifting and compressing the thresholded response in an enumerative fashion. Third column: For each source of variability, the plausible responses were re-arranged into vectors, which were subsequently stacked into a data matrix (variability matrix). Fourth column: The eigenvalue plots show the explained variance for each of the first 20 generated PCs, for each variability matrix. Note that the first two PCs explain the largest part of the total variance of each source of variability. Fifth column: Five regressors capturing the average magnitude of the considered TF-feature, and its variability in latency (the second PC in latency-shift variability matrix), in frequency (the second PC in frequency-shift variability matrix), in morphology in the time domain (the second PC in latency-compression variability matrix), and in morphology in the frequency domain (the second PC in frequency-compression variability matrix).
Mentions: To obtain an accurate estimate of single trial time-frequency responses, not only their variability in latency and frequency, but also their variability in morphology (both in time and frequency domain) should be taken into account. This has a physiological rationale. For example, in some clinical conditions (e.g., optic neuritis during multiple sclerosis), visual-evoked potentials are “desynchronized”, i.e., their amplitudes are reduced because of increased latency jitter, as well as increased width of single-trial responses (Orssaud, 2003; Pelosi et al., 1997). Latency jitter, as well as trial-by-trial variability in response morphology (i.e., wave width or frequency variability) could thus be important parameters for clinical studies. Therefore, in addition to the basis set of TF-MLR, two more regressors, representing the scaling of the single-trial response in time or frequency domain are considered, which leads to the following TF-MLRd model (Fig. 4):(4)Ftf=k1F1s1t+a1c1f+b1+k2F2s2t+a2c1f+b2+k3F3s3t+a3c1f+b3+εwhere s1, s2 and s3 are the coefficients that determine the compression ratios of the time width of ERP, ERD and ERS of each single-trial TFD compared to those of the average TFD, respectively, while c1, c2 and c3 are the coefficients that determine the compression ratios of the frequency width of ERP, ERD and ERS of each single-trial TFD compared to those of the average TFD, respectively.

Bottom Line: Transient sensory, motor or cognitive event elicit not only phase-locked event-related potentials (ERPs) in the ongoing electroencephalogram (EEG), but also induce non-phase-locked modulations of ongoing EEG oscillations.ERD and ERS reflect changes in the parameters that control oscillations in neuronal networks and, depending on the frequency at which they occur, represent neuronal mechanisms involved in cortical activation, inhibition and binding.This permits within-subject statistical comparisons, correlation with pre-stimulus features, and integration of simultaneously-recorded EEG and fMRI.

View Article: PubMed Central - PubMed

Affiliation: Key Laboratory of Cognition and Personality (Ministry of Education) and Faculty of Psychology, Southwest University, Chongqing, China; Department of Neuroscience, Physiology and Pharmacology, University College London, UK. Electronic address: huli@swu.edu.cn.

No MeSH data available.


Related in: MedlinePlus