An fMRI Study to find aligning Network to the signals related to the Decision making process
Poster No:
1792
Submission Type:
Abstract Submission
Authors:
Dayoung Yoon1, Su Hyun Bong2, Jae Won Kim3, Bumseok Jeong4
Institutions:
1KAIST, Daejeon, Daejeon, 2KAIST, E7 building lobby, Daejeon, 3Seoul National University College of Medicine, Jongno-gu, Seoul, 4Korea Advanced Institute of Science and Technology, Daejeon, Daejeon
First Author:
Co-Author(s):
Introduction:
In the extensive study of brain regions crucial for the decision-making process,various behavioral models have been proposed to elucidate their interconnected roles.Notably, dopamine,considered a pivotal neurotransmitter,has garnered significant attention.The prevailing argument suggests that dopamine signals encode the disparity in prediction errors during the observation and inference processes.However the Active Inference, one of the most influential theories in decision-making and action control introduces a novel perspective posits that dopaminergic firing specifically encodes the precision of actions.We aims to scrutinize the validity of this hypothesis in light of the connectivity.
Methods:
The fMRI data was collected from 58 participants while they are performing two-armed bandit reinforcement learning task with Pavlovian-instrumental transfer(PIT) paradigm.The task consists of two parts whether subjects get or lose the reward by their choice(Fig1).Participants need to learn which arm is better to get the reward or avoid loss. After preprocessing, we acquire time series from cortex with Schaefer Atlas[1] and subcortex with Melbourne Atlas[2],and to include other regions known to be related to the release of Dopamine and Norepinephrine,we also add some regions in CIT168[3] and Harvard Ascending Arousal Network Atlas[4].From this, we derived 6 types of timeseries data, each comprising a 3 by 2 structure.3 denotes the hierarchical levels of where the signals are generated, encompassing each node, each edge, and each network.And 2 signifies that one aspect captures the signal changes at each level while the other encapsulates the dynamic changes of intersubject correlation(ISC)[5].At the node level, the signals correspond to the activation of individual brain regions.Moving to the edge level, these signals represent the connectivity between two nodes and are calculated using phase-locked oscillatory pattern obtained through LEiDA[6].Finally, at the network level, to estimate the representative networks, we emloyed the graph Laplacian mixture Model[7] and yielded the four most representative brain states. Next, we constructed Active Inference model of two armed bandit task that is neural process theories within the Bayesian framework [8] and fit it to behavioral data of choices and outcomes.As a result we can get two signals from each participants.One is the rate of changes of expected precision value, and the other one is state prediction error. We convolved hemodynamic response function to both signals, and compared them to the above 6 types of neuroimaging data using generalized linear model with both convolved signals as independent variables and neuroimaing data as dependent variables.The coefficients values related to each signal that are derived from the edge level are compared using Network-Based Statistic[9].And we also performed one-samle test using null distribution which are generated by phase randomization of both signals to coefficients values that are from the other two levels.
Results:
Averaged 6 type of timeseries are shown in Fig1.Four extracted states and fractional occupancy of each state are shown in Fig3.The results of one sample test are shown in Table2.Significant nodes and states are indicated.At node level, both signals shares numerous common regions but certain nodes are specific to one of the signals.At the network level,State1,2 are common to both signals,but State3 in the Precision signal is distinctive.At the edge level, the differences are shown in Fig4.While the nodes involved in these connections overlap with those exhibiting significance in the node-level,the connections themselves manifest distinct patterns in both signals.
·Table2. NA; Nodal Activation, SSPE; Signed State Prediction Error, dISC; dynamic intersubject correlation, dISC_st; dynamic intersubject correlation of state, P; Precision, S; SSPE
·Fig4. a,c)Prec>SSPE in edge connectivity and dISFC. b,d)Prec
Conclusions:
Despite some overlap due to the intimate association between the two signals,we can identify distinct connectivity patterns associated with each signal. and showed that these differences at the edge level were also reflected in the difference in the network level explained by State3.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI)
Bayesian Modeling 2
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 1
Keywords:
Dopamine
FUNCTIONAL MRI
Modeling
1|2Indicates the priority used for review
Provide references using author date format
[2]Tian, Y., Margulies, D. S., Breakspear, M., & Zalesky, A. (2020). Topographic organization of the human subcortex unveiled with functional connectivity gradients. Nature Neuroscience, 23(11), 1421–1432
[3]Pauli, W. M., Nili, A. N., & Michael Tyszka, J. (2018). Data Descriptor: A high-resolution probabilistic in vivo atlas of human subcortical brain nuclei. Scientific Data, 5.
[4]Edlow, B. L., Takahashi, E., Wu, O., Benner, T., Dai, G., Bu, L., … Folkerth, R. D. (2012). Neuroanatomic connectivity of the human ascending arousal system critical to consciousness and its disorders. Journal of Neuropathology and Experimental Neurology, 71(6), 531–546
[5]Hasson, U., Ghazanfar, A. A., Galantucci, B., Garrod, S., & Keysers, C. (2012). Brain-to-brain coupling: a mechanism for creating and sharing a social world. Trends in Cognitive Sciences, 16(2), 114–121
[6]Vohryzek, J., Deco, G., Cessac, B., Kringelbach, M. L., & Cabral, J. (2020). Ghost Attractors in Spontaneous Brain Activity: Recurrent Excursions Into Functionally-Relevant BOLD Phase-Locking States. Frontiers in Systems Neuroscience, 14.
[7]Ricchi, I., Tarun, A., Maretic, H. P., Frossard, P., & Van De Ville, D. (2022). Dynamics of functional network organization through graph mixture learning. NeuroImage, 252.
[8]Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., & Pezzulo, G. (2017). Active inference: A process theory. Neural Computation, 29(1), 1–49
[9]Zalesky, A., Fornito, A., & Bullmore, E. T. (2010). Network-based statistic: Identifying differences in brain networks. NeuroImage, 53(4), 1197–1207