Advertisement
Research Article

Cortical Modulations Increase in Early Sessions with Brain-Machine Interface

  • Miriam Zacksenhouse mail,

    To whom correspondence should be addressed. E-mail: mermz@tx.technion.ac.il

    Affiliation: Faculty of Mechanical Engineering, Technion, Haifa, Israel

    X
  • Mikhail A. Lebedev,

    Affiliations: Department of Neurobiology, Duke University, Durham, North Carolina, United States of America, Center for Neuro-engineering, Duke University, Durham, North Carolina, United States of America

    X
  • Jose M. Carmena,

    Affiliations: Department of Neurobiology, Duke University, Durham, North Carolina, United States of America, Center for Neuro-engineering, Duke University, Durham, North Carolina, United States of America

    X
  • Joseph E. O'Doherty,

    Affiliation: Department of Biomedical Engineering, Duke University, Durham, North Carolina, United States of America

    X
  • Craig Henriquez,

    Affiliations: Center for Neuro-engineering, Duke University, Durham, North Carolina, United States of America, Department of Biomedical Engineering, Duke University, Durham, North Carolina, United States of America

    X
  • Miguel A.L. Nicolelis

    Affiliations: Department of Neurobiology, Duke University, Durham, North Carolina, United States of America, Center for Neuro-engineering, Duke University, Durham, North Carolina, United States of America, Department of Biomedical Engineering, Duke University, Durham, North Carolina, United States of America, Department of Psychological and Brain Sciences, Duke University, Durham, North Carolina, United States of America

    X
  • Published: July 18, 2007
  • DOI: 10.1371/journal.pone.0000619

Reader Comments (1)

Post a new comment on this article

Referee Comments: Referee #2

Posted by PLoS_ONE_Group on 30 Jul 2007 at 02:03 GMT

The authors present a sophisticated analysis of the variance in neural responses obtained from several areas of sensorimotor cortex of macaque monkeys as they learn to use neural activity to control a brain-machine interface. The main result is clear: early in the process the variance is high and it reduces as performance improves. The authors speculate that error processing or explorative activity accounts for the early burst of activity modulation. The results are new, and well documented. A disappointment for me is that results were similar for different cortical areas, so not much could be inferred about the distinct roles of these areas. We now know more about the learning process, but it is not clear to me what the next step is. I have no suggestions for change.

N.B. These are the general comments made by the reviewer when reviewing this paper in light of which the manuscript was revised. Specific points addressed during revision of the paper are not shown.


RE: Referee Comments: Referee #2

MiriamZ replied to PLoS_ONE_Group on 07 Aug 2007 at 19:57 GMT

The comment raises two important issues, which are addressed below: (1) Distinction between different cortical areas, and (2) Possible implications.

(1) Distinction between different cortical areas: Indeed all the recorded neural areas (M1, PMd, S1, SMA and PP) exhibited higher neural modulations during brain control than during pole control. However, the magnitude and significance of the change in neural modulations did depend on the cortical region. In particular (see Figure 7 and Table 2): PMd neurons exhibited low modulations in pole control and the largest relative change after the onset of brain control; M1 neurons exhibited high modulations even in pole control and the largest absolute change; and SMA neurons exhibited a very small change. These region-dependent differences agree well with our suggested interpretation that the enhanced neural modulations involve error representation and corrective activity, as detailed in the Discussion. PMd neurons, for example, are reported to be involved in error correction. Error correction is expected to be minor during the highly skilled pole control but should become significant in the novel task associated with brain control via the BMI – in agreement with our results.

(2) Implications and future work: The next step is to characterize the signals encoded in the enhanced modulations. This step is important both for understanding sensori-motor control and learning, and for evaluating potential BMI improvements. Assuming that our interpretation is correct, the enhanced neural modulations represent movement errors and required movement corrections. Under this interpretation, current BMIs extract only the desired velocity, which is generated by the feed-forward path, while ignoring the corrective velocity command, which is generated in the feedback path. By characterizing the representation of movement corrections, it should be possible to extract this information too and use it to improve the operation of the BMI.