Predicting metacognitive abilities across sensory modalities with functional networks
Poster No:
934
Submission Type:
Abstract Submission
Authors:
Katarzyna Hat1, Paola Galdi2, Kristian Sandberg3, Michał Wierzchoń4
Institutions:
1Consciousness Lab, Psychology Institute, & Centre for Brain Research, Jagiellonian University, Kraków, Lesser Poland, 2School of Informatics, University of Edinburgh, Edinburgh, United Kingdom, 3Center of Functionally Integrative Neuroscience, Aarhus University, Aarhus, Denmark, 4Consciousness Lab, Psychology Institute, & Centre for Brain Research, Jagiellonian University, Kraków, Poland
First Author:
Katarzyna Hat
Consciousness Lab, Psychology Institute, & Centre for Brain Research, Jagiellonian University
Kraków, Lesser Poland
Consciousness Lab, Psychology Institute, & Centre for Brain Research, Jagiellonian University
Kraków, Lesser Poland
Co-Author(s):
Kristian Sandberg
Center of Functionally Integrative Neuroscience, Aarhus University
Aarhus, Denmark
Center of Functionally Integrative Neuroscience, Aarhus University
Aarhus, Denmark
Michał Wierzchoń
Consciousness Lab, Psychology Institute, & Centre for Brain Research, Jagiellonian University
Kraków, Poland
Consciousness Lab, Psychology Institute, & Centre for Brain Research, Jagiellonian University
Kraków, Poland
Introduction:
Metacognition is a person's ability to correctly assess and control their cognitive processes. One of the leading debates in metacognition research is whether it is a domain-general or a domain-specific process (Rouault et al., 2018). Here, we investigate metacognition in different sensory modalities with resting-state networks.
Methods:
Behavioural data: 4 homologous perceptual tasks were administered outside the scanner in 4 sensory modalities: vision, audition, nociception and touch. Each task consisted of 2 alternative forced choice tasks followed by confidence judgement on a 10%-step size scale from 50% (guessing) to 100% (fully confident). Data was analysed with the Signal Detection Framework using the bhsdtr2 package (Paulewicz & Blaut, 2020), and a type-2 d' (metacognitive sensitivity - meta-d') was estimated (Maniscalco & Lau, 2012).
Resting-state fMRI: 302 participants underwent resting-state fMRI recording on a Siemens Skyra 3T (eyes open with fixation cross; TR=801ms, voxel size=2,5mm isotropic, 18min of acquisition). Data were preprocessed with fMRIPrep 21.0.2 and rsDenoise (github) following Finn et al. (2015), excluding Global Signal Regression.
Functional connectivity: Denoised data were parcellated into 400 cortical parcels from Schaefer et al. (2018), 17 subcortical parcels defined as in the HCP CIFTI files (Glasser et al., 2013), and 28 cerebellar parcels using the SUIT atlas (Diedrichsen et al., 2009). Average time series were extracted from parcels to estimate functional connectivity as pairwise Pearson's correlation.
Predictive framework: Functional connections were used to predict the meta-d' scores using elastic net linear regression and nested cross-validation. We included data from 193 subjects in visual, 182 in auditory, 135 in tactile and 161 in nociceptive task. We run predictions using all connections (whole-brain) and different combinations of artificial lesions, following Dubois et al. (2018), to measure the impact of specific resting-state networks on prediction. We used the 17 Yeo networks (2011) plus the cerebellum and subcortical regions as additional networks, for a total of 19 networks. We used 'all but 1' and 'all but 2' networks lesions (i.e. using only 1 or 2 networks for prediction). We measured Pearson's correlation between predicted and actual meta-d' scores and controlled for false discovery rate (FDR) with the Benjamini-Hochberg procedure, setting the significance level at α=0.05.
Resting-state fMRI: 302 participants underwent resting-state fMRI recording on a Siemens Skyra 3T (eyes open with fixation cross; TR=801ms, voxel size=2,5mm isotropic, 18min of acquisition). Data were preprocessed with fMRIPrep 21.0.2 and rsDenoise (github) following Finn et al. (2015), excluding Global Signal Regression.
Functional connectivity: Denoised data were parcellated into 400 cortical parcels from Schaefer et al. (2018), 17 subcortical parcels defined as in the HCP CIFTI files (Glasser et al., 2013), and 28 cerebellar parcels using the SUIT atlas (Diedrichsen et al., 2009). Average time series were extracted from parcels to estimate functional connectivity as pairwise Pearson's correlation.
Predictive framework: Functional connections were used to predict the meta-d' scores using elastic net linear regression and nested cross-validation. We included data from 193 subjects in visual, 182 in auditory, 135 in tactile and 161 in nociceptive task. We run predictions using all connections (whole-brain) and different combinations of artificial lesions, following Dubois et al. (2018), to measure the impact of specific resting-state networks on prediction. We used the 17 Yeo networks (2011) plus the cerebellum and subcortical regions as additional networks, for a total of 19 networks. We used 'all but 1' and 'all but 2' networks lesions (i.e. using only 1 or 2 networks for prediction). We measured Pearson's correlation between predicted and actual meta-d' scores and controlled for false discovery rate (FDR) with the Benjamini-Hochberg procedure, setting the significance level at α=0.05.
Results:
We found significant results for both types of lesions but not for whole-brain predictions. For brevity, we focus only on 'all but 2 networks' lesions. The functional connections selected as most predictive for each task are shown in Figure 1. All pairs of networks that produced significant predictions are listed in Table 1 together with the correlation scores and associated q-values (p-values after FDR correction). Most of the pairs are specific to a given modality but some pairs of networks appear in multiple modalities, and some networks by themself are prominent in all or most modalities.
·Figure 1. Predictive edges for each modality obtained with 'all but 2 networks' lesions.
·Table 1. Predictive pairs of networks per modality with q-values.
Conclusions:
Our analysis suggests that metacognition can be studied with a 'lesion' approach, while the mechanism is not robust enough to be detected in the whole brain analysis, at least with the current sample. We found each modality portrays a different set of predictive pairs of networks, suggesting at least some part of processing to be domain-specific. However, we also found candidates for metacognitive hubs, especially control A and ventral attention B networks, which are present in results for all modalities. Also default B, somatomotor B and dorsal attention B have the potential to participate in more general processes of metacognition, with the first two not present only in visual modality, and default B participating in multiple pairings. Given the data acquisition was run with eyes open and fixation cross, stronger activation of visual regions could potentially impact the predictivity of metacognitive abilities in this modality.
Higher Cognitive Functions:
Executive Function, Cognitive Control and Decision Making 1
Higher Cognitive Functions Other
Perception, Attention and Motor Behavior:
Consciousness and Awareness 2
Perception: Multisensory and Crossmodal
Perception: Tactile/Somatosensory
Keywords:
Consciousness
Meta-Cognition
MRI
Perception
Somatosensory
1|2Indicates the priority used for review
Provide references using author date format
Dubois, J. (2018). A distributed brain network predicts general intelligence from resting-state human neuroimaging data. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1756), 20170284.
Finn, E. S. (2015). Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nature neuroscience, 18(11), 1664–1671.
Glasser, M. F. (2013). The minimal preprocessing pipelines for the Human Connectome Project. NeuroImage, 80, 105–124.
Maniscalco, B. (2012). A signal detection theoretic approach for estimating metacognitive sensitivity from confidence ratings. Consciousness and cognition, 21(1), 422–430. https://doi.org/10.1016/j.concog.2011.09.021.
Paulewicz, B. (2020). The bhsdtr package: A general-purpose method of Bayesian inference for signal detection theory models. Behavior Research Methods, 52(5), 2122–2141.
Rouault, M. (2018). Human metacognition across domains: insights from individual differences and neuroimaging. Personality neuroscience, 1, e17. https://doi.org/10.1017/pen.2018.16.
rsDenoise: https://github.com/adolphslab/rsDenoise
Schaefer, A. (2018). Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. Cerebral cortex (New York, N.Y. : 1991), 28(9), 3095–3114.
Yeo, B. T. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of neurophysiology, 106(3), 1125–1165.