Decision letter | Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation

Open accessCopyright infoDownload PDFDownload figuresRelated content

Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation

Decision letter

Affiliation details

City College of the City University of New York, United States; New York University School of Medicine, United States; Mayo Clinic, United States
Richard Ivry, Reviewing editor, University of California, Berkeley, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor, Richard Ivry, and Sabine Kastner as the Senior Editor. The following individual involved in review of your submission has agreed to reveal his identity: Angel Peterchev (Reviewer #1).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

Thank you for submitting your work entitled "Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation" for consideration at eLife. As one of the reviewers noted, "This manuscript is of exceptional importance as it provides unique, much needed, and highly sophisticated measurement, modeling, and analysis of the electric field in the human brain due to transcranial electrical stimulation (tES)." The paper provides the first thorough validation of computational models of transcranial current stimulation. While there are concerns with the modeling approaches used to make electric field estimates, the measured fields in the vicinity of the electrodes provide a direct measure that are independent of the modeling. These estimates provide the most direct observations of tES induced E-field in the human brain, an extremely important contribution. The data set is substantial and, as indicated by the authors, will be made available publicly, a valuable resource for the community. In summary, the reviewers felt that the paper, when appropriately revised, will be a landmark report in this literature and of great use to the growing community interested in using, modeling, and optimizing this stimulation technology.

Essential revisions:

I) Concerns with the experimental results: These experimental results are very important. It is essential to make sure that the methods and analysis are clearly described and consider possible confounds, especially factors that might contribute to signal attenuation.

1) Along these lines, we want the revision to specify the input impedance of the amplifiers and the typical electrode impedance. It is stated that electrodes were considered high impedance if there was strong 60 Hz noise on these channels, but it is unclear if this approach eliminated electrodes with low enough impedance to prevent 60 Hz noise, but high enough impedance to cause an attenuation of the recorded voltages. Obviously, any unintended voltage attenuation, while not critical for intracranial EEG, can affect all estimates of E-field strength and tissue conductivities in this study. Furthermore, it is implied that there was clipping of the signal in some electrodes (data from them were not included in the analysis). During clipping did the input impedance of the recording amplifier decrease and could this produce current stimulus flow though the recording electrodes?

2) There is a preprint available on bioaRxiv that involves similar experiments on two patients and two NHP (http://biorxiv.org/content/early/2016/05/18/053892, Opitz et al.). The results reported in that preliminary report could be compared to the present work. Given the great interest we anticipate in your work and this other study, it would be useful to comment on the similarities and differences between the experimental results in two studies. Two issues that come up when comparing the studies are 1) possible frequency-dependent effects and the manner in which these were analyzed, and 2) differences in how electric fields are estimated. For example, for electrodes in the interior of the grid why not use all neighbors instead of just the closest one?

II) Concerns with the modeling work: We see the modeling work as a nice complement to the experimental data. We recognize that this work entails complex models, involving a number of assumptions. Below we summarize major issues to consider in revising this section. Most involve clarifying the methods and potential confounds of the models, and carefully qualifying the conclusions based on these models. In some cases, we would recommend conducting additional simulations and/or analysis since this will reduce qualifications you would have to place on the conclusions.

3) The study concludes that modeling anatomical detail such as the skull layers and the white matter anisotropy does not improve the model predictions. It seems, however, that the models were optimized only for the "realistic model" without the respective anatomical refinements. This should be clarified, and if this is indeed the case, the comparisons are biased since one side of the comparison is optimized and the other is not. Furthermore, the white matter conductivity anisotropy affects mostly the E-field in the white matter, while most of the voltage recordings were not in the white matter, which could affect efforts to optimize the anisotropic model fit to the data.

4) (also relevant to experimental results) The selected montage (e.g., inferior position), combined with the large inter-electrode distance increases the significance of conduction paths in the lower half of the head. These include regions with complex anatomy and conductivity profiles such as the eyes, orbits, optic nerve, neck, nasal cavity, etc., as well as the truncation of the head and neck (as seen in Figure 4). Furthermore, only a limited set of tissues are modeled, and other potentially significant ones are omitted such as muscle, fat, and sub-compartments of the eye. Discussion should be provided concerning the extent these regions affect the accuracy of the electric field simulation as well as the derived "optimal" conductivity of the scalp and skull.

5) The derived "optimal" conductivity values for scalp (mean = 3.61 V/m) are very high, being an order of magnitude above values in the literature and more than twice the measured conductivity of CSF and physiological saline (both ~ 1.6 S/m). It is unclear what physical mechanism could possibly confer such high conductivity. While this is acknowledged by the authors, this is an issue of concern. Since they are based on a fit of a complicated model with just two free parameters to a limited data set, it is entirely possible that the optimized conductivity values result from unrecognized limitations of the model and/or recordings. For example, truncation of the head and neck (as seen in Figure 4) or other anatomical inaccuracies in the lower portion of the head could artificially strengthen the modelled current flow in the top portion of the head, requiring higher scalp conductivity values to match the recorded potentials. Further, the optimized conductivity values may depend on the location of the tES electrodes. Adopting the "optimal" values from this study in general for tES (and potentially other) simulations may produce errors when the model context changes. We would like you to consider quantifying the impact of these model limitations; for example, test the impact of head truncation and reduction of the number of segmented tissue compartments in one of the full head and neck models. In addition, the revision should be explicit about these limitations and impact on inferences to be drawn from this work.

6) The paper attempts to solve two mathematical problems: (A) An inverse problem in which they the estimate conductivity (scalp and skull only) given the voltage measurements noted above and (B) A forward problem in which they calculate the scalar potential field and the electric field everywhere (or at least potentially so) within the head given the previously estimated conductivity and assumptions of their model (6 regions delineated by anatomical MRI with each region being of spatially constant conductivity). The solution of problem (A) is attempted through a minimization of error between measured and predicted values as given by Equation 1. The solution of problem (B) is obtained through the Abaqus software. For the estimates of the electric field to be meaningful, a reasonable accuracy estimate must be associated with the estimate. No such estimate is given. To do so would require at least the following: (i) A reasonable estimate of the error produced in results of problems (A) and (B) that is associated with assumption that the conductivity is spatially constant throughout each of the 6 segmented regions and (ii) A reasonable estimate of the error in the results of problems (A) and (B) associated the measurement and segmentation of the six regions. In addition, known instrumental error must be propagated through both problems as well.

7) Why aren't the correlation values between measured and predicted fields better than are reported here? If the solution to problem (A) is unique and if the voltage measurements are consistent with the quasi-static field equations, then why don't the results of problem (B) reproduce the measurements of voltages (albeit with some presumably small measurement error discrepancies)?

III) Concerns with the organization and clarity of the paper:

8) In general, there is concern with organization, clarity, precision, and to some extent discussion. The current organization is hard to follow. There are methods descriptions mixed into the Results section, while at the same time parameters reported on in Results (e.g. "s", "r", "t(n)" – the latter does not even seem to be explained in Methods)) without any description at that point of what they mean or any mention that they are described in Methods. Please review your manuscript from the reader point of view, rigorously stripping methods material from Results and inserting descriptive phrases and referrals to Methods in Results where needed. The statistics should be described more carefully in the Methods section. A number of experimental details are missing or not clear, including the specific criteria by which some patients were excluded and also by which electrodes were excluded within a given patient (since the latter is a very significant fraction of the total number of electrodes this seems particularly important) and what the range of electrodes was across all subjects (we are only given the total). Make sure the manuscript clearly describes when and why particular sections/graphs exclude some of the participants. There are a number of places were the grammar needs work, including instances where there is disagreement of subject and verb, missing prepositions, missing definite or indefinite articles, etc. Please review the manuscript to be as precise as possible (e.g., what does "as important as commonly thought" in the subsection “Relative merit of various model refinements”?).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation" for further consideration at eLife. Your revised article has been favorably evaluated by Sabine Kastner as the Senior Editor, Rich Ivry as the Reviewing Editor, and two reviewers.

In general, we are pleased with the extensive revisions you have provided. We continue to believe this is a very important paper, one that will be of considerable interest to the brain stimulation community.

There remains one important issue to be addressed. While you have provided E field predictions, the manuscript does not provide an associated estimate of the error term for these predictions. A complete response would address the following:

1) For conductivities that are allowed to vary in the current fitting procedure: Provide the range of conductivity values that work nearly as well as those used in the current version.

2) For conductivities that have been fixed in the fitting procedure: Provide a range of conductivity values that fall within reasonable estimates of error for those values.

3) A range of compartment segmentations taking into account a reasonable estimate of error for the accuracy of the segmentation methods.

We recognize that the work required for these three error estimates may be extensive, and also see considerable value in the current modeling work given that the electrical field estimates for each neighboring electrode contact pair correlate well with the measurements, indicating that the shape of the model solution is in agreement with the shape of the data. Moreover, the conductivities are in reasonable agreement with published values.

That said, at a minimum we would like to see #1 addressed, with a quantification of the uncertainty (confidence intervals) of the optimized conductivity values. Are the confidence intervals for the individually estimated conductivities available? This would provide information on the uncertainty in the optimization process, assuming the non-optimized parameters are fixed. In addition, you could propagate reasonable estimates of error in the model parameters through the predictive model to estimate error in the prediction.

We also ask that you consider #2 and #3, and in the ideal response, would again propogate the error estimates. The former could be done by varying the fixed conductivity parameters. As we see it, #3, while useful, might entail a lot of work and the added value would not justify the request. Nonetheless, if you have information of value here, this would certainly strengthen the modeling component of the paper. If you decide to focus on #1 only, then we would ask that, in the revision, you make explicit limitations with the current work in terms of establishing the accuracy of your approach since the paper would lack a thorough analysis of the estimates of error.

We apologize for the long turnaround time on this revision, but wanted to provide a clear request for what we anticipate will be a final revision, one that we should be able to act on in a very timely manner.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation" for further consideration at eLife. Your revised article has been favorably evaluated by Sabine Kastner as the Senior Editor, Rich Ivry as the Reviewing Editor, and two reviewers.

Just two revision requests to consider. I am confident that you will be able to address them quickly and then we can move to a final decision after the Reviewing Editor has a chance to take a look at the revision.

1) Reviewer 1 has two suggestions for improving Figure 5—figure supplement 1:a) In Figure 8—figure supplement 1A–C, logarithmic spacing of the y axis could be considered given the wide range of the values (over a decade) and the seeming proportional behavior of the C-R bound error bars.b) In Figure 8—figure supplement 1D–E, add lines connecting the two points for each subject in the corresponding color to allow easier visual inspection of the local sensitivity.

2) Reviewer 2 asks for more on your efforts to provide estimates of error. I am satisfied with what you have done on this. You could consider compiling the discussion of error estimation into a focused section to highlight this issue for readers. I will leave the final decision here up to you.

DOI: http://dx.doi.org/10.7554/eLife.18834.016