Limits...
The role of ECoG magnitude and phase in decoding position, velocity, and acceleration during continuous motor behavior.

Hammer J, Fischer J, Ruescher J, Schulze-Bonhage A, Aertsen A, Ball T - Front Neurosci (2013)

Bottom Line: The decoding accuracy based only on phase information was substantially (at least 2 fold) and significantly higher than that based only on magnitudes for position, velocity and acceleration.There was also no indication of additional magnitude-based mechanisms encoding movement information in the LFC range.Thus, our study contributes to elucidating the nature of the informative LFC of motor cortical population activity and may hence contribute to improve decoding strategies and BMI performance.

View Article: PubMed Central - PubMed

Affiliation: Bernstein Center Freiburg, University of Freiburg Freiburg, Germany.

ABSTRACT
In neuronal population signals, including the electroencephalogram (EEG) and electrocorticogram (ECoG), the low-frequency component (LFC) is particularly informative about motor behavior and can be used for decoding movement parameters for brain-machine interface (BMI) applications. An idea previously expressed, but as of yet not quantitatively tested, is that it is the LFC phase that is the main source of decodable information. To test this issue, we analyzed human ECoG recorded during a game-like, one-dimensional, continuous motor task with a novel decoding method suitable for unfolding magnitude and phase explicitly into a complex-valued, time-frequency signal representation, enabling quantification of the decodable information within the temporal, spatial and frequency domains and allowing disambiguation of the phase contribution from that of the spectral magnitude. The decoding accuracy based only on phase information was substantially (at least 2 fold) and significantly higher than that based only on magnitudes for position, velocity and acceleration. The frequency profile of movement-related information in the ECoG data matched well with the frequency profile expected when assuming a close time-domain correlate of movement velocity in the ECoG, e.g., a (noisy) "copy" of hand velocity. No such match was observed with the frequency profiles expected when assuming a copy of either hand position or acceleration. There was also no indication of additional magnitude-based mechanisms encoding movement information in the LFC range. Thus, our study contributes to elucidating the nature of the informative LFC of motor cortical population activity and may hence contribute to improve decoding strategies and BMI performance.

No MeSH data available.


Related in: MedlinePlus

Decoding of velocity from different signal components and anatomical areas. Results for subjects S1–S3. (A) All grid channels decoding. Mean CC ± s.e.m. as a function of time offset τ between ECoG predictors and decoded movement velocity (negative values of τ indicating that the ECoG preceeds in time the movement execution). The 4 different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), time-frequency magnitude-only (yellow), and time domain LFC (cyan). Decoding accuracy of time-frequency magnitude and phase features at their peak values (labeled with red triangles) are significantly better than those of the time domain LFC. (B)Time-frequency magnitude and phase decoding using single channels and time offset τ indicated by red triangles in (A). The square plots represent the ECoG grid of each subject, with marked central (and in S1 lateral) sulcus (thick white curve on the grid), division of anatomical areas (thin white lines, cf., Figures 2E–G) and the labeled electrical stimulation results (label color: magenta—motor response, gray—sensory response. H—hand, A—arm, O—oral, E—eyes, L—leg, S—shoulder, N—neck). (C) Decoding of channel groups based on assignment to anatomical areas (section Electrical Stimulation Mapping and Channel Assignment to Anatomical Areas in main text) using the time-frequency magnitude and phase features. Colors of the anatomical areas are the same as those in Figures 2E–G. Premotor area (in blue) provides the most accurate predictions in all subjects. (D) Decoding from relative power modulations in a wide range of different frequency bands. The correlation coefficient is color-coded for each frequency band (f1 − f2), as defined by the x- and y-axes. Results are shown for the time offset τ with the maximal correlation in each subjects (0.0 s, −0.1 s, and 0.2 s for subjects S1–S3, respectively). The phase of the slow oscillations was clearly more informative than any of the spectral band power features.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3814578&req=5

Figure 7: Decoding of velocity from different signal components and anatomical areas. Results for subjects S1–S3. (A) All grid channels decoding. Mean CC ± s.e.m. as a function of time offset τ between ECoG predictors and decoded movement velocity (negative values of τ indicating that the ECoG preceeds in time the movement execution). The 4 different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), time-frequency magnitude-only (yellow), and time domain LFC (cyan). Decoding accuracy of time-frequency magnitude and phase features at their peak values (labeled with red triangles) are significantly better than those of the time domain LFC. (B)Time-frequency magnitude and phase decoding using single channels and time offset τ indicated by red triangles in (A). The square plots represent the ECoG grid of each subject, with marked central (and in S1 lateral) sulcus (thick white curve on the grid), division of anatomical areas (thin white lines, cf., Figures 2E–G) and the labeled electrical stimulation results (label color: magenta—motor response, gray—sensory response. H—hand, A—arm, O—oral, E—eyes, L—leg, S—shoulder, N—neck). (C) Decoding of channel groups based on assignment to anatomical areas (section Electrical Stimulation Mapping and Channel Assignment to Anatomical Areas in main text) using the time-frequency magnitude and phase features. Colors of the anatomical areas are the same as those in Figures 2E–G. Premotor area (in blue) provides the most accurate predictions in all subjects. (D) Decoding from relative power modulations in a wide range of different frequency bands. The correlation coefficient is color-coded for each frequency band (f1 − f2), as defined by the x- and y-axes. Results are shown for the time offset τ with the maximal correlation in each subjects (0.0 s, −0.1 s, and 0.2 s for subjects S1–S3, respectively). The phase of the slow oscillations was clearly more informative than any of the spectral band power features.

Mentions: Using the decoding approach based on a time-frequency representation of the ECoG as described in section Fourier Descriptors of Short-Time Fourier transform, we first addressed the question whether LFC phase is indeed the major source of movement-related decodable information. By comparing results from DAlg5 (magnitude information only) and DAlg4 (phase information only) obtained from the same data set, it was possible to quantitatively assess the relative contributions of magnitude and phase separately. We compared the decoding performance taking all channels of the ECoG grid at one specific time offset τ to predict the velocity. This time offset τ was systematically varied over the interval [−3.5, 3.5] s. Note that we defined the time offset τ such that its negative value corresponds to the situation where ECoG activity precedes the movement (see Methods, section Decoding model, for further details). We found that in all subjects phase clearly proved to be substantially more informative than magnitude in all trajectory derivatives (Figures 6C,D, for direct comparison of velocity prediction in single subjects cf., yellow and magenta curves in Figure 7A). Peak correlation coefficients (CCs) between actual and predicted velocity for all subjects movement validation folds were in the range 0.46 ± 0.10 (mean ± std) for phase-only features, which was significantly higher (paired, two-sided sign test of 30 CCs from each cross-validation fold and time lag, P = 0.001 significance level, false discovery rate correction for multiple tests over time lags) than for magnitude-only features, where CC = 0.16 ± 0.12 (Table 2). Moreover, maximal DA achieved using phase only was very similar to that obtained using the time-domain LFC (DAlg1, Figure 6A, Table 2, also cf., magenta and cyan curves in Figure 7A). These findings clearly identify phase (and not the magnitude) as the major carrier of information for ECoG LFC decoding of movement velocity.


The role of ECoG magnitude and phase in decoding position, velocity, and acceleration during continuous motor behavior.

Hammer J, Fischer J, Ruescher J, Schulze-Bonhage A, Aertsen A, Ball T - Front Neurosci (2013)

Decoding of velocity from different signal components and anatomical areas. Results for subjects S1–S3. (A) All grid channels decoding. Mean CC ± s.e.m. as a function of time offset τ between ECoG predictors and decoded movement velocity (negative values of τ indicating that the ECoG preceeds in time the movement execution). The 4 different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), time-frequency magnitude-only (yellow), and time domain LFC (cyan). Decoding accuracy of time-frequency magnitude and phase features at their peak values (labeled with red triangles) are significantly better than those of the time domain LFC. (B)Time-frequency magnitude and phase decoding using single channels and time offset τ indicated by red triangles in (A). The square plots represent the ECoG grid of each subject, with marked central (and in S1 lateral) sulcus (thick white curve on the grid), division of anatomical areas (thin white lines, cf., Figures 2E–G) and the labeled electrical stimulation results (label color: magenta—motor response, gray—sensory response. H—hand, A—arm, O—oral, E—eyes, L—leg, S—shoulder, N—neck). (C) Decoding of channel groups based on assignment to anatomical areas (section Electrical Stimulation Mapping and Channel Assignment to Anatomical Areas in main text) using the time-frequency magnitude and phase features. Colors of the anatomical areas are the same as those in Figures 2E–G. Premotor area (in blue) provides the most accurate predictions in all subjects. (D) Decoding from relative power modulations in a wide range of different frequency bands. The correlation coefficient is color-coded for each frequency band (f1 − f2), as defined by the x- and y-axes. Results are shown for the time offset τ with the maximal correlation in each subjects (0.0 s, −0.1 s, and 0.2 s for subjects S1–S3, respectively). The phase of the slow oscillations was clearly more informative than any of the spectral band power features.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3814578&req=5

Figure 7: Decoding of velocity from different signal components and anatomical areas. Results for subjects S1–S3. (A) All grid channels decoding. Mean CC ± s.e.m. as a function of time offset τ between ECoG predictors and decoded movement velocity (negative values of τ indicating that the ECoG preceeds in time the movement execution). The 4 different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), time-frequency magnitude-only (yellow), and time domain LFC (cyan). Decoding accuracy of time-frequency magnitude and phase features at their peak values (labeled with red triangles) are significantly better than those of the time domain LFC. (B)Time-frequency magnitude and phase decoding using single channels and time offset τ indicated by red triangles in (A). The square plots represent the ECoG grid of each subject, with marked central (and in S1 lateral) sulcus (thick white curve on the grid), division of anatomical areas (thin white lines, cf., Figures 2E–G) and the labeled electrical stimulation results (label color: magenta—motor response, gray—sensory response. H—hand, A—arm, O—oral, E—eyes, L—leg, S—shoulder, N—neck). (C) Decoding of channel groups based on assignment to anatomical areas (section Electrical Stimulation Mapping and Channel Assignment to Anatomical Areas in main text) using the time-frequency magnitude and phase features. Colors of the anatomical areas are the same as those in Figures 2E–G. Premotor area (in blue) provides the most accurate predictions in all subjects. (D) Decoding from relative power modulations in a wide range of different frequency bands. The correlation coefficient is color-coded for each frequency band (f1 − f2), as defined by the x- and y-axes. Results are shown for the time offset τ with the maximal correlation in each subjects (0.0 s, −0.1 s, and 0.2 s for subjects S1–S3, respectively). The phase of the slow oscillations was clearly more informative than any of the spectral band power features.
Mentions: Using the decoding approach based on a time-frequency representation of the ECoG as described in section Fourier Descriptors of Short-Time Fourier transform, we first addressed the question whether LFC phase is indeed the major source of movement-related decodable information. By comparing results from DAlg5 (magnitude information only) and DAlg4 (phase information only) obtained from the same data set, it was possible to quantitatively assess the relative contributions of magnitude and phase separately. We compared the decoding performance taking all channels of the ECoG grid at one specific time offset τ to predict the velocity. This time offset τ was systematically varied over the interval [−3.5, 3.5] s. Note that we defined the time offset τ such that its negative value corresponds to the situation where ECoG activity precedes the movement (see Methods, section Decoding model, for further details). We found that in all subjects phase clearly proved to be substantially more informative than magnitude in all trajectory derivatives (Figures 6C,D, for direct comparison of velocity prediction in single subjects cf., yellow and magenta curves in Figure 7A). Peak correlation coefficients (CCs) between actual and predicted velocity for all subjects movement validation folds were in the range 0.46 ± 0.10 (mean ± std) for phase-only features, which was significantly higher (paired, two-sided sign test of 30 CCs from each cross-validation fold and time lag, P = 0.001 significance level, false discovery rate correction for multiple tests over time lags) than for magnitude-only features, where CC = 0.16 ± 0.12 (Table 2). Moreover, maximal DA achieved using phase only was very similar to that obtained using the time-domain LFC (DAlg1, Figure 6A, Table 2, also cf., magenta and cyan curves in Figure 7A). These findings clearly identify phase (and not the magnitude) as the major carrier of information for ECoG LFC decoding of movement velocity.

Bottom Line: The decoding accuracy based only on phase information was substantially (at least 2 fold) and significantly higher than that based only on magnitudes for position, velocity and acceleration.There was also no indication of additional magnitude-based mechanisms encoding movement information in the LFC range.Thus, our study contributes to elucidating the nature of the informative LFC of motor cortical population activity and may hence contribute to improve decoding strategies and BMI performance.

View Article: PubMed Central - PubMed

Affiliation: Bernstein Center Freiburg, University of Freiburg Freiburg, Germany.

ABSTRACT
In neuronal population signals, including the electroencephalogram (EEG) and electrocorticogram (ECoG), the low-frequency component (LFC) is particularly informative about motor behavior and can be used for decoding movement parameters for brain-machine interface (BMI) applications. An idea previously expressed, but as of yet not quantitatively tested, is that it is the LFC phase that is the main source of decodable information. To test this issue, we analyzed human ECoG recorded during a game-like, one-dimensional, continuous motor task with a novel decoding method suitable for unfolding magnitude and phase explicitly into a complex-valued, time-frequency signal representation, enabling quantification of the decodable information within the temporal, spatial and frequency domains and allowing disambiguation of the phase contribution from that of the spectral magnitude. The decoding accuracy based only on phase information was substantially (at least 2 fold) and significantly higher than that based only on magnitudes for position, velocity and acceleration. The frequency profile of movement-related information in the ECoG data matched well with the frequency profile expected when assuming a close time-domain correlate of movement velocity in the ECoG, e.g., a (noisy) "copy" of hand velocity. No such match was observed with the frequency profiles expected when assuming a copy of either hand position or acceleration. There was also no indication of additional magnitude-based mechanisms encoding movement information in the LFC range. Thus, our study contributes to elucidating the nature of the informative LFC of motor cortical population activity and may hence contribute to improve decoding strategies and BMI performance.

No MeSH data available.


Related in: MedlinePlus