Frequently Asked Questions#
Is this service free to use?
Yes! Note, however, that Neuroscout is a web-based engine for fMRI analysis specification; at the moment, we donât provide free computing resources for the execution of the resulting analysis bundles. Analyses can be run on Google Colabâa demo notebook is provided here.
I plan to publish results Iâve obtained using Neuroscout. How do I cite it?
After you generate an analysis, a âBibliographyâ tab will be shown which will auto-generate a reference list for the dataset, feature extractors, and scientific software used for that analysis. In addition to these references, be sure to include the unique analysis ID associated with any results.
Also, be sure to cite the Neuroscout platform as follows:
Neuroscout, a unified platform for generalizable and reproducible fMRI research. Alejandro de la Vega, Roberta Rocca, Ross W. Blair, Christopher J. Markiewicz, Jeff Mentch, James D. Kent, Peer Herholz, Satrajit S. Ghosh, Russell A. Poldrack, Tal Yarkoni. bioRxiv 2022.04.05.487222; doi: https://doi.org/10.1101/2022.04.05.487222
Are there any restrictions on analyses Iâve created?
Yes. By using Neuroscout, you agree that once you have finalized and âcompiledâ an analysis, the analysis can no longer be deleted from our system. If you wish to âeditâ an analysis, you may clone the analysis, and make any desired changed on the forked version. Although analyses are by default not searchable by other users, any user with the private analysis ID may view your analysis.
Also, in the event that you publish any results generated using the NeuroScout interface, you MUST provide a link to the corresponding analysis page(s) on the NeuroScout website.
I have a naturalistic study Iâd like to share on Neuroscout. How do I do so?
Due to the cost of extracting features from multi-modal stimuli using external APIs, the set of datasets we support is manually curated. However, we are continually expanding the list of supported datasets, and we encourage researchers to contact us if they want to make their data available for use in Neuroscout. Note that it is much easier for us to ingest datasets that are already deposited in the OpenNeuro repository, and we we strongly recommend uploading your dataset to OpenNeuro whether or not it eventually ends up in Neuroscout.
I want to make changes to an analysis I already ran, but it is locked. How can I edit it?
Once an analysis has been run, it is permanently locked and archived for provenance. You may âcloneâ your analysis, and make changes to this new copy of your analysis. This iterative workflow ensures the provenance of existing analyses, while allowing users the flexibility to continue to modify and improve models.
I want to make one of my âprivateâ analyses public, but the website says the analysis is âlockedâ!
When an analysis is locked, you can no longer make any substantive changes that affect model specification. However, you can always edit the name, description, and public/private status. So go ahead and make your analysis public!
How do you automatically extract features from naturalistic datasets?
The original stimuli presented to users are submitted to various machine learning algorithms and services to extract novel feature timecourses. To facility this process, we have developed a Python library for multimodal feature extraction called pliers. Pliers allows us to extract a wide-variety of features across modalities using various external content analysis services with ease. For example, we are able to use Google Vision API to encode various aspects of the visual elements of movie frames, such as when a face is present.
In addition, pliers allows us to easily link up various feature extraction services; for example, we can use the IBM Watson Speech to Text API to transcribe the speech in a movie into words with precise onsets, and then use a predefined dictionary of lexical norms to extract lexical norms for each word, such as frequency. We can then generate timecourses for each of these extracted features, creating novel predictors of brain activity.
For more information of pliers, please see the GitHub repository and the following paper:
McNamara, Q., De La Vega, A., & Yarkoni, T. (2017, August). Developing a comprehensive framework for multimodal feature extraction. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 1567-1574). ACM.
Am I restricted to mass univariate GLMs, or can I use Neuroscout to specify other kinds of analyses?
Currently, that is the primary analysis mode of Neuroscout. However, the underlying BIDS-StatsModel is designed with more complex models in mind, such as predictive and linear-mixed effect models.
In addition, using pyNS, users are free to access the Neuroscout API and develop novel analysis workflows using extracted Predictors and Datasets. The Neuroscout team is also actively expanding the types of models that can be created to including predictive encoding and decoding multivariate models.
Can I contribute my own predictors to Neuroscout?
Yes! Using the âMy Predictorsâ function, you can create custom collections of predictors to add to your analyses. Simply navigate to My Predictors and click on âAdd New Predictorsâ.
For more information, see the guide on Upload Predictors