Review for "AFQ-Browser: Supporting reproducible human neuroscience research through browser-based visualization tools"

Completed on 27 Nov 2017 by Roey Schurr .

Login to endorse this review.


This paper presents AFQ-Browser, a new brweser-based tool for sharing, visualizing and exploring neuroimaging (as well as demographic and behavioral) data based on the AFQ analysis pipeline. This tool is designed to allow interactive data exploration, and makes extensive use of linked visualization options, to allow a deeper understanding of rich multidimensional data. The manuscript also presents some insightful examples of how the AFQ Browser can be used for making new discoveries on published data sets. The paper fills an ever growing need in the neuroscience community, to be able to replicate published results and easily share big data.

Comments to author

- The first two paragraphs in the introductions could benefit from moving some of the technical details further down.

- Linked visualization (page 6): The visualization of actual fiber bundles, and especially their link to the profiles figures is a wonderful feature. However, it should be noted that the shown bundles are not taken from the data itself, but rather from a template or representative subject. It is therefore important to stress that one has to make sure the tractography itself was successful, particularly for diseased populations, where the tractography itself might be problematic. It is also important since the tracts’ profiles might differ due to different tractography results (e.g. results with a lot of crossing between tracts A and B could explains a certain decrease in FA, whereas the template tractography may show no crossing between the two bundles).

- Page 8: The example of how changes in the CST may lead to a decrease in FA is a great example, with a lot of pedagogical value.

- Page 10: In Figure 3 you write “Mean diffusivity (top panel) and radial diffusivity (middle panel) show larger group differences than fractional anisotropy (bottom panel)”. What exactly do you mean by “larger” group differences? Is the difference measured by the number of healthy-populaiton-SD?

- Page 11: “Even though a multivariate classification strategy (random forests) is used to achieve good diagnostic accuracy, visualization of individual Tract Profiles demonstrates that a majority of patients (75%) deviated by more than 1SD from control values within the right CST at the level of the cerebral peduncle.” – Could you please elaborate? Are you saying that the classification doesn’t give us insight regarding the exact location of the difference, which can be gained by visualizing the data?

- Page 13: “While there are a few regions that show small differences (depending on the statistical threshold)…” – Could you please clarify, why do you say that the differences depend on statistical threshold?