Print Close

Impact of Individualized Parcellation on State and Network Specific Brain-Behavior Relationships

Poster No:

1749

Submission Type:

Abstract Submission

Authors:

Nevena Kraljević¹^,2, Robert Langner¹^,2, Federico Raimondo¹^,2, Kaustubh Patil¹^,2, Ru Kong³, Leon Ooi³, B. T. Thomas Yeo³, Simon Eickhoff¹^,2, Veronika Müller¹^,2

Institutions:

¹Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf, Düsseldorf, Germany, ²Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich, Jülich, Germany, ³National University of Singapore, Singapore, Singapore

First Author:

Nevena Kraljević
Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Düsseldorf, Germany|Jülich, Germany

Co-Author(s):

Robert Langner
Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Düsseldorf, Germany|Jülich, Germany

Federico Raimondo
Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Düsseldorf, Germany|Jülich, Germany

Kaustubh Patil
Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Düsseldorf, Germany|Jülich, Germany

Ruby Kong
National University of Singapore
Singapore, Singapore

Leon Ooi
National University of Singapore
Singapore, Singapore

B. T. Thomas Yeo
National University of Singapore
Singapore, Singapore

Simon Eickhoff
Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Düsseldorf, Germany|Jülich, Germany

Veronika Müller
Institute of Systems Neuroscience, Medical Faculty, HHU Düsseldorf|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Düsseldorf, Germany|Jülich, Germany

E-Poster

Introduction:

Behavior prediction based on brain data can contribute to our understanding of human brain functioning and even to personalized neuroscience. Various approaches have been developed to improve brain-behavior prediction: using task-based functional connectivity (FC) rather than resting-state FC [1]; using feature-reduction methods [2–5]; or individualizing brain features [4-6]. In our previous work, we explored how the correspondence of fMRI state and network priors with predicted target behavior (i.e. state and network specificity) influences behavioral prediction. Despite overall weak predictions, we observed a tendency for state specificity in working memory (WM), but none for theory-of-mind cognition (SOCIAL) or emotion matching (EMO). Employing an algorithm to individualize functional networks [6], we here assessed systematically if stronger brain-behavior relationships for states and networks that fit with the predicted behavioral domain can be found when using individualized brain features for prediction.

Methods:

We used openly available behavioral and fMRI data from the Human Connectome Project - Young Adult S1200 [7] (340 participants, 186 women, mean age = 28.7 years) dataset. A multi-session hierarchical Bayesian model was employed on resting-state data, factoring in group functional and spatial network information, to estimate individual-specific cortical network parcellations [6]. To obtain the FC features, minimally preprocessed fMRI data was used regressing out age, sex and motion. FC matrices were generated for two parcellation schemes (non-individualized vs. individualized), and two connectome representations (whole brain 400-Schaefer atlas [8] vs. three task-related networks [9]) each from four task fMRI states (resting/WM/SOCIAL/EMO). In a machine-learning approach applying partial least squares regression, the FC matrices were used as features to predict (z-transformed) performance of three tasks performed in the scanner: WM, SOCIAL, and EMO. Prediction performance was evaluated by coefficient of determination (COD) and mean root mean squared error (RMSE) between predicted and observed score and from the cross-validation.

Results:

Task states predicted WM performance better than resting state (lower RMSE, see Fig 1A), while for SOCIAL and EMO it was the other way around (Fig. 1B-C). Whole-brain FC patterns predicted WM and EMO a bit better than did task-based networks (Fig. 2). However, none of those differences were significant. Generally, the prediction models showed a poor fit with low prediction performance (COD of 0.17). Prior individualization didn't lead to better prediction performance and did in particular not reveal stronger state and network specificity as none of the parcellation schemes (individualized, non-individualized) indicated a benefit of network correspondence.

Conclusions:

In conclusion, predicting behavior based on FC remains a significant challenge. However, in line with previous results [1] task-specific FC patterns seem to provide more information about individual behavior in WM performance than do resting-state FC patterns but this does not generalize to socio-affective behaviors. Calculation of FC within individualized networks did not enhance the expected state and network specificity, indicating that the use of individual functional markers does not necessarily capture more task-specific variance than do nodes from group-averaged maps. However, this might be due to the rather small amount of participants and the fact that the individualization was based on resting-state data [6], while individualization on task-fMRI data specific to the target domain might improve prediction and reveal specific effects of state and network. Notably, given the limitations and the generally low prediction accuracies, the observed absence of differences in prediction performance between state and network conditions as well as the lack of impact of individualization need to be viewed with caution and invite further investigations.

Emotion, Motivation and Social Neuroscience:

Social Neuroscience Other

Learning and Memory:

Working Memory

Modeling and Analysis Methods:

Classification and Predictive Modeling ²

fMRI Connectivity and Network Modeling ¹

Task-Independent and Resting-State Analysis

Keywords:

Cognition

Computational Neuroscience

FUNCTIONAL MRI

Machine Learning

Social Interactions

Other - Prediction

^1|2Indicates the priority used for review

Provide references using author date format

1. Finn, E. S. 'Is it time to put rest to rest?', Trends in Cognitive Sciences, 25, 1021–1032 (2021).
2. Finn, E. S. et al. 'Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity', Nature Neuroscience, 18, 1664–1671 (2015).
3. Pläschke, R. N. et al. 'On the integrity of functional brain networks in schizophrenia, Parkinson’s disease, and advanced age: Evidence from connectivity-based single-subject classification', Human Brain Mapping, 38, 5845–5858 (2017).
4. Shen, X. et al. Using connectome-based predictive modeling to predict individual behavior from brain connectivity. Nat. Protoc. 12, 506–518 (2017).
5. Chen, J. et al. 'Intrinsic Connectivity Patterns of Task-Defined Brain Networks Allow Individual Prediction of Cognitive Symptom Dimension of Schizophrenia and Are Linked to Molecular Architecture', Biological Psychiatry, 89, 308–319 (2021).
6. Kong, R. et al. 'Spatial Topography of Individual-Specific Cortical Networks Predicts Human Cognition, Personality, and Emotion', Cerebral Cortex, 29, 2533–2551 (2019).
7. Van Essen, D. C. et al. 'The WU-Minn Human Connectome Project: An overview', NeuroImage, 80, 62–79 (2013).
8. Schaefer, A. et al. 'Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI', Cerebral Cortex, 28, 3095–3114 (2018).
9. Kraljević, N. et al. 'Network and State Specificity in Connectivity-Based Predictions of Individual Behavior', http://biorxiv.org/lookup/doi/10.1101/2023.05.11.540387 (2023).

The work was supported by: Helmholtz Portfolio Theme "Supercomputing and Modeling for the Human Brain“; European Union’s Horizon 2020 Research and Innovation Programme under Grant Agreement No. 945539 (HBP SGA3); fellowship of the German Academic Exchange Service (DAAD).