Poster No:
1112
Submission Type:
Abstract Submission
Authors:
Nils Winter1, Ramona Leenings1, Lukas Fisch2, Jan Ernsting1, Carlotta Barkhau1, Maximilian Konowski3, Udo Dannlowski4, Tim Hahn1
Institutions:
1University of Münster, Münster, Germany, 2Institute for Translational Psychiatry, Münster, North Rhine Westphalia, 3University of Münster, Münster, Nordrhein-Westfalen, 4Institute for Translational Psychiatry, Münster, North Rhine-Westphalia
First Author:
Co-Author(s):
Lukas Fisch
Institute for Translational Psychiatry
Münster, North Rhine Westphalia
Udo Dannlowski
Institute for Translational Psychiatry
Münster, North Rhine-Westphalia
Tim Hahn
University of Münster
Münster, Germany
Introduction:
Connectome-based predictive modeling (CPM) is an increasingly popular approach for gaining insights into the neurobiological underpinnings of mental disorders and individual differences (Ju et al., 2020; Shen et al., 2017). A critical challenge in these analyses is the control of confounding variables-factors that influence both the brain connectome and the outcomes of interest.(Snoek et al., 2019) If left uncontrolled, confounds can cause misleading, spurious conclusions about brain-behavior associations. However, there is significant heterogeneity in the methods currently used in the literature to control for confounding effects (Dinga et al., 2020; Rosenblatt et al., 2024). Some studies do not address confounds at all, while others rely on partial correlations during connectome edge selection. Another common method involves regressing out confounds from the brain connectome data, which may not fully eliminate the confounding effect (Dinga et al., 2020). Currently, a systematic evaluation of these methods is lacking, and no comprehensive toolbox exists that offers multiple approaches to managing confounds in CPM analyses. In response to these challenges, we evaluate existing methods and introduce a straightforward and intuitive approach utilizing predictive model increments. Furthermore, we offer a robust, Python-based software tool for CPM analyses, incorporating confound correction.
Methods:
To evaluate different confound correction methods, we simulated data including connectome, confound, and outcome of interest, with varying degrees of direct and indirect associations. Our proposed method, using predictive model increments, involves building a baseline model with only confounding variables to predict the desired outcome. We then construct a second model that includes both CPM brain network strength and confounds. The difference in predictive accuracy-termed the predictive increment-assesses the unique contribution of the connectome to the outcome prediction. Similar to standard CPM analyses, permutation tests are conducted to derive empirical null distributions of predictive increments, providing a non-parametric measure of statistical significance for the unique contributions of brain data to the predictive model.
Results:
Our experiments with simulated data show that predictive increments effectively identify the unique contribution of connectome data in predictive modeling. Even when confounds strongly influenced both the connectome and outcomes, our method reliably discerned the genuine predictive power of the connectome, independently of confounding variables. To facilitate effective confound control in CPM, we developed a publicly available Python toolbox (https://wwu-mmll.github.io/confound_corrected_cpm/). This toolbox implements CPM analyses and includes methods for controlling confounds, optimizing edge selection, and enhancing model building. It provides researchers with a practical tool to manage confounding variables, thereby improving the accessibility and applicability of CPM analyses across the research community.
Conclusions:
Predictive increments provide an intuitive and efficient approach to managing confounding variables in CPM studies, enabling researchers to attribute predictive utility specifically to connectome data. This method addresses common methodological issues, leading to more reliable neurobiological insights. Importantly, our Python toolbox offers a comprehensive framework for CPM analyses, integrating methods for handling confounds. This toolbox sets new standards in connectome research by ensuring observed associations genuinely reflect underlying brain-behavior associations.
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
Connectivity (eg. functional, effective, structural) 2
fMRI Connectivity and Network Modeling
Methods Development
Keywords:
Computational Neuroscience
Data analysis
FUNCTIONAL MRI
Machine Learning
MRI
Open-Source Software
Statistical Methods
1|2Indicates the priority used for review
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.
I accept
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:
I do not want to participate in the reproducibility challenge.
Please indicate below if your study was a "resting state" or "task-activation” study.
Other
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Healthy subjects
Was this research conducted in the United States?
No
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.
Not applicable
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.
Not applicable
Please indicate which methods were used in your research:
Computational modeling
For human MRI, what field strength scanner do you use?
3.0T
Which processing packages did you use for your study?
Other, Please list
Provide references using APA citation style.
for effects of confounding variables on machine learning predictions. https://doi.org/10.1101/2020.08.17.255034
Ju, Y., Horien, C., Chen, W., Guo, W., Lu, X., Sun, J., Dong, Q., Liu, B., Liu, J., Yan, D., Wang, M., Zhang, L., Guo, H., Zhao, F., Zhang, Y., Shen, X., Constable, R. T., & Li, L. (2020). Connectome-based models can predict early symptom improvement in major depressive disorder. Journal of Affective Disorders, 273, 442–452. https://doi.org/10.1016/j.jad.2020.04.028
Rosenblatt, M., Tejavibulya, L., Jiang, R., Noble, S., & Scheinost, D. (2024). Data leakage inflates prediction performance in connectome-based machine learning models. Nature Communications, 15(1), 1829. https://doi.org/10.1038/s41467-024-46150-w
Shen, X., Finn, E. S., Scheinost, D., Rosenberg, M. D., Chun, M. M., Papademetris, X., & Constable, R. T. (2017). Using connectome-based predictive modeling to predict individual behavior from brain connectivity. Nature Protocols, 12(3), 506–518. https://doi.org/10.1038/nprot.2016.178
Snoek, L., Miletić, S., & Scholte, H. S. (2019). How to control for confounds in decoding analyses of neuroimaging data. NeuroImage, 184, 741–760. https://doi.org/10.1016/j.neuroimage.2018.09.074
No