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

Single channel velocity decoding and tuning analysis. Columns represent the three selected channels from subjects S1–S3, respectively. (A) Single-channel decoding accuracy at different time offset τ (same notation as in Figure 5A). The three different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), and time-domain LFC (cyan). The selected channel corresponds to 1st row and 3rd column in the ECoG grid of subject S1 (cf., Figure 2). Note the “camel back” shape of the LFC-based decoding with two separate peaks indicated by “1” and “2.” (B) Velocity—ECoG LFC tuning. The x-axis defines the time offset τ between velocity and ECoG data, the y-axis defines the velocity bins v (where v > 0 for rightward movements and v < 0 for leftward movements). The binned ECoG LFC average is color-coded (and interpolated). Around τ = 0 s, a polarity gradient of mean LFC from positive (for v < 0) to negative (for v > 0) can be observed (left star and square, respectively). Approximately 1 s later, the polarity was opposite (right square and star). The minimum of LFC-based decoding accuracy in (A) as indicated by a vertical dotted line clearly corresponds to the time of polarity reversal in (B), where the LFC showed little tuning. (C) Velocity—ECoG FDf tuning, where f = 0.5 Hz. The velocity binning is identical as in (B), the complex-valued FDf response is transparency-color-coded (scaled as indicated by the color bar in the lower-right corner), where transparency indicates the magnitude of the averaged FDf features and their phase angles, defined as arctg(Re/Im), are coded by a circular color map. At the time points of minimal LFC decoding accuracy (vertical dotted lines), the phase is still tuned, due to the polarity changes as shown in (B), explaining the high phase-based DA at these time points (magenta curve in A). (D–F) same as in (A–C), the selected channel corresponds to 5th row and 4th column of the ECoG grid of subject S2. (G–I) same as in (A–C), the selected channel corresponds to 8th row and 5th column of the ECoG grid of subject S3.
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 10: Single channel velocity decoding and tuning analysis. Columns represent the three selected channels from subjects S1–S3, respectively. (A) Single-channel decoding accuracy at different time offset τ (same notation as in Figure 5A). The three different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), and time-domain LFC (cyan). The selected channel corresponds to 1st row and 3rd column in the ECoG grid of subject S1 (cf., Figure 2). Note the “camel back” shape of the LFC-based decoding with two separate peaks indicated by “1” and “2.” (B) Velocity—ECoG LFC tuning. The x-axis defines the time offset τ between velocity and ECoG data, the y-axis defines the velocity bins v (where v > 0 for rightward movements and v < 0 for leftward movements). The binned ECoG LFC average is color-coded (and interpolated). Around τ = 0 s, a polarity gradient of mean LFC from positive (for v < 0) to negative (for v > 0) can be observed (left star and square, respectively). Approximately 1 s later, the polarity was opposite (right square and star). The minimum of LFC-based decoding accuracy in (A) as indicated by a vertical dotted line clearly corresponds to the time of polarity reversal in (B), where the LFC showed little tuning. (C) Velocity—ECoG FDf tuning, where f = 0.5 Hz. The velocity binning is identical as in (B), the complex-valued FDf response is transparency-color-coded (scaled as indicated by the color bar in the lower-right corner), where transparency indicates the magnitude of the averaged FDf features and their phase angles, defined as arctg(Re/Im), are coded by a circular color map. At the time points of minimal LFC decoding accuracy (vertical dotted lines), the phase is still tuned, due to the polarity changes as shown in (B), explaining the high phase-based DA at these time points (magenta curve in A). (D–F) same as in (A–C), the selected channel corresponds to 5th row and 4th column of the ECoG grid of subject S2. (G–I) same as in (A–C), the selected channel corresponds to 8th row and 5th column of the ECoG grid of subject S3.

Mentions: To further explore the role of phase in motor decoding, we performed single channel decoding and tuning analyses. In the analyses based on all channels as described above, decoding utilizing phase showed a smoother time course than LFC-based decoding (see above, Figures 6, 7A). This difference was even more pronounced at the single channel level (Figure 10). The time offset course of LFC-based decoding typically showed clearly distinct, multiple peaks, while decoding based on phase (alone or in combination with magnitude) was much smoother (cyan vs. red/magenta curves in Figure 10A). This effect can be intuitively understood from time-resolved single channel velocity tuning of the different signal components.


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)

Single channel velocity decoding and tuning analysis. Columns represent the three selected channels from subjects S1–S3, respectively. (A) Single-channel decoding accuracy at different time offset τ (same notation as in Figure 5A). The three different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), and time-domain LFC (cyan). The selected channel corresponds to 1st row and 3rd column in the ECoG grid of subject S1 (cf., Figure 2). Note the “camel back” shape of the LFC-based decoding with two separate peaks indicated by “1” and “2.” (B) Velocity—ECoG LFC tuning. The x-axis defines the time offset τ between velocity and ECoG data, the y-axis defines the velocity bins v (where v > 0 for rightward movements and v < 0 for leftward movements). The binned ECoG LFC average is color-coded (and interpolated). Around τ = 0 s, a polarity gradient of mean LFC from positive (for v < 0) to negative (for v > 0) can be observed (left star and square, respectively). Approximately 1 s later, the polarity was opposite (right square and star). The minimum of LFC-based decoding accuracy in (A) as indicated by a vertical dotted line clearly corresponds to the time of polarity reversal in (B), where the LFC showed little tuning. (C) Velocity—ECoG FDf tuning, where f = 0.5 Hz. The velocity binning is identical as in (B), the complex-valued FDf response is transparency-color-coded (scaled as indicated by the color bar in the lower-right corner), where transparency indicates the magnitude of the averaged FDf features and their phase angles, defined as arctg(Re/Im), are coded by a circular color map. At the time points of minimal LFC decoding accuracy (vertical dotted lines), the phase is still tuned, due to the polarity changes as shown in (B), explaining the high phase-based DA at these time points (magenta curve in A). (D–F) same as in (A–C), the selected channel corresponds to 5th row and 4th column of the ECoG grid of subject S2. (G–I) same as in (A–C), the selected channel corresponds to 8th row and 5th column of the ECoG grid of subject S3.
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 10: Single channel velocity decoding and tuning analysis. Columns represent the three selected channels from subjects S1–S3, respectively. (A) Single-channel decoding accuracy at different time offset τ (same notation as in Figure 5A). The three different features analyzed are: time-frequency magnitude and phase (red), time-frequency phase-only (magenta), and time-domain LFC (cyan). The selected channel corresponds to 1st row and 3rd column in the ECoG grid of subject S1 (cf., Figure 2). Note the “camel back” shape of the LFC-based decoding with two separate peaks indicated by “1” and “2.” (B) Velocity—ECoG LFC tuning. The x-axis defines the time offset τ between velocity and ECoG data, the y-axis defines the velocity bins v (where v > 0 for rightward movements and v < 0 for leftward movements). The binned ECoG LFC average is color-coded (and interpolated). Around τ = 0 s, a polarity gradient of mean LFC from positive (for v < 0) to negative (for v > 0) can be observed (left star and square, respectively). Approximately 1 s later, the polarity was opposite (right square and star). The minimum of LFC-based decoding accuracy in (A) as indicated by a vertical dotted line clearly corresponds to the time of polarity reversal in (B), where the LFC showed little tuning. (C) Velocity—ECoG FDf tuning, where f = 0.5 Hz. The velocity binning is identical as in (B), the complex-valued FDf response is transparency-color-coded (scaled as indicated by the color bar in the lower-right corner), where transparency indicates the magnitude of the averaged FDf features and their phase angles, defined as arctg(Re/Im), are coded by a circular color map. At the time points of minimal LFC decoding accuracy (vertical dotted lines), the phase is still tuned, due to the polarity changes as shown in (B), explaining the high phase-based DA at these time points (magenta curve in A). (D–F) same as in (A–C), the selected channel corresponds to 5th row and 4th column of the ECoG grid of subject S2. (G–I) same as in (A–C), the selected channel corresponds to 8th row and 5th column of the ECoG grid of subject S3.
Mentions: To further explore the role of phase in motor decoding, we performed single channel decoding and tuning analyses. In the analyses based on all channels as described above, decoding utilizing phase showed a smoother time course than LFC-based decoding (see above, Figures 6, 7A). This difference was even more pronounced at the single channel level (Figure 10). The time offset course of LFC-based decoding typically showed clearly distinct, multiple peaks, while decoding based on phase (alone or in combination with magnitude) was much smoother (cyan vs. red/magenta curves in Figure 10A). This effect can be intuitively understood from time-resolved single channel velocity tuning of the different signal components.

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