Representational connectivity analyses to track information across the brain
Poster No:
1270
Submission Type:
Late-Breaking Abstract Submission
Authors:
SEPIDEH KILANI1, Jason Mattingley2, Hamid Karimi-Rouzbahani3
Institutions:
1Queensland Brain Institute, Brisbane, Australia, 2University of Queensland, St Lucia, Queensland, 3Queensland Brain Institute, The University of Queensland, Brisbane, Queensland
First Author:
Co-Author(s):
Hamid Karimi-Rouzbahani
Queensland Brain Institute, The University of Queensland
Brisbane, Queensland
Queensland Brain Institute, The University of Queensland
Brisbane, Queensland
Late Breaking Reviewer(s):
Ruby Kong
Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor
Singapore, Singapore
Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor
Singapore, Singapore
Introduction:
The brain's computational power arises from both individual neuronal activity and complex neuronal interactions (Anzellotti et al., 2018). Cognitive neuroscience often seeks to understand how experimental manipulations are represented in the brain and how representations move across the brain. To address this, we and others have developed Representational Connectivity Analyses (RCA) to measure how neural representations distinguishing cognitive states are transferred between brain areas (Fig 1).
There are two main categories of RCA: model-free which extracts connectivity patterns directly from data without assumptions about the propagated information (Goddard et al., 2016; Kietzmann et al., 2019), and model-based methods which track specific aspects of information using predefined models (Karimi-Rouzbahani et al., 2021). Previously, we provided a conceptual comparison of these methods (Karimi-Rouzbahani et al., 2022). Here, we implement and systematically assess two model-free, and one model-based RCA method across several simulated scenarios.
There are two main categories of RCA: model-free which extracts connectivity patterns directly from data without assumptions about the propagated information (Goddard et al., 2016; Kietzmann et al., 2019), and model-based methods which track specific aspects of information using predefined models (Karimi-Rouzbahani et al., 2021). Previously, we provided a conceptual comparison of these methods (Karimi-Rouzbahani et al., 2022). Here, we implement and systematically assess two model-free, and one model-based RCA method across several simulated scenarios.
Methods:
We generated activity patterns for two ROIs across 16 experimental cognitive conditions using MATLAB RSA toolbox (Nili et al., 2014). Specifically, we simulated three scenarios:
1) We assessed how common noise affects connectivity estimates in model-free and model-based approaches. We generated Gaussian noise (zero-mean) with different variances (Fig 2) for the first ROI. We then applied a random transformation matrix (zero-mean, variance of 10) to introduce correlated patterns of activity across ROIs.
2) We tested if choosing a multi-component model for model-based methods can lead to spurious connectivity, which does not happen when using model-free methods. To that end, we generated two independent ROIs that encoded distinct types of information and used a two-component model which contained aspects of both types of information represented in the two ROIs.
3) We simulated two ROIs coding distinct types of information, but, here, the two information types were encoded across the ROIs with congruent timing: the destination ROI was delayed relative to the source ROI, which suggested the transformation of information from the source to the destination. We then determined whether using a model-based approach with relevant models for each ROI can detect information transfer (connectivity) unaffected by its transformation.
1) We assessed how common noise affects connectivity estimates in model-free and model-based approaches. We generated Gaussian noise (zero-mean) with different variances (Fig 2) for the first ROI. We then applied a random transformation matrix (zero-mean, variance of 10) to introduce correlated patterns of activity across ROIs.
2) We tested if choosing a multi-component model for model-based methods can lead to spurious connectivity, which does not happen when using model-free methods. To that end, we generated two independent ROIs that encoded distinct types of information and used a two-component model which contained aspects of both types of information represented in the two ROIs.
3) We simulated two ROIs coding distinct types of information, but, here, the two information types were encoded across the ROIs with congruent timing: the destination ROI was delayed relative to the source ROI, which suggested the transformation of information from the source to the destination. We then determined whether using a model-based approach with relevant models for each ROI can detect information transfer (connectivity) unaffected by its transformation.
Results:
1) When common noise was added, it spuriously inflated connectivity between ROIs in the model-free methods, while there was little change in connectivity in the model-based method (Fig 2). This suggests that while the model-free methods can be affected by common noise, model-based methods are less affected as they target specific aspects rather than the shared noise.
2) In the scenario with two independent ROIs encoding unrelated information, connectivity assessed via a multi-component model showed spurious connectivity using model-based RCA. In contrast, the model-free RCA analysis correctly indicated no connectivity. This highlights that inappropriate choice of model in model-based RCA can lead to false-positive findings.
3) We find that while the model-free RCAs missed connectivity as a result of transformed patterns of information, the model-based analysis successfully detected connectivity across the ROIs.
2) In the scenario with two independent ROIs encoding unrelated information, connectivity assessed via a multi-component model showed spurious connectivity using model-based RCA. In contrast, the model-free RCA analysis correctly indicated no connectivity. This highlights that inappropriate choice of model in model-based RCA can lead to false-positive findings.
3) We find that while the model-free RCAs missed connectivity as a result of transformed patterns of information, the model-based analysis successfully detected connectivity across the ROIs.
Conclusions:
Our findings highlight the importance of selecting appropriate methods for assessing neural connectivity, especially when tracking information across the brain in cognitive experiments. Model-free RCA is sensitive to shared noise, leading to spurious connectivity, while model-based RCA can introduce false positives if the model assumptions are incorrect. On the other hand, model-based methods are capable of detecting connectivity when information transforms from one area to another, whereas model-free methods may miss that transformed information. Therefore, careful method selection and validation are crucial for accurately tracking information flow in the brain.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural) 1
Multivariate Approaches 2
Keywords:
Cognition
Computational Neuroscience
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
Provide references using APA citation style.
Goddard, E., Carlson, T. A., Dermody, N., & Woolgar, A. (2016). Representational dynamics of object recognition: Feedforward and feedback information flows. Neuroimage, 128, 385-397.
Karimi-Rouzbahani, H., Ramezani, F., Woolgar, A., Rich, A., & Ghodrati, M. (2021). Perceptual difficulty modulates the direction of information flow in familiar face recognition. NeuroImage, 233, 117896.
Karimi-Rouzbahani, H., Woolgar, A., Henson, R., & Nili, H. (2022). Caveats and nuances of model-based and model-free representational connectivity analysis. Frontiers in Neuroscience, 16, 755988.
Kietzmann, T. C., Spoerer, C. J., Sörensen, L. K., Cichy, R. M., Hauk, O., & Kriegeskorte, N. (2019). Recurrence is required to capture the representational dynamics of the human visual system. Proceedings of the National Academy of Sciences, 116(43), 21854-21863.
Nili, H., Wingfield, C., Walther, A., Su, L., Marslen-Wilson, W., & Kriegeskorte, N. (2014). A toolbox for representational similarity analysis. PLoS computational biology, 10(4), e1003553.