% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@INPROCEEDINGS{PaasOliveros:1014932,
author = {Paas Oliveros, Lya Katarina and Jung, Kyesam and Hu, Dan
and Müller, Veronika and Heckner, Marisa and Pläschke,
Rachel and Eickhoff, Simon and Liu, Hesheng and Langner,
Robert},
title = {{P}redicting dual-task performance from individualized
functional and structural netwo},
reportid = {FZJ-2023-03487},
year = {2023},
note = {Acknowledgments: This study was supported by the Deutsche
Forschungsgemeinschaft (DFG, LA 3071/3-1, SFB 1451), the
National Institute of Mental Health (R01-MH074457), the
Helmholtz Portfolio Theme Supercomputing and Modeling for
the Human Brain, and the European Union’s Horizon 2020
Research and Innovation Program under Grant Agreement No.
945539 (HBP SGA3), and a German Academic Exchange Service
(DAAD) Scholarship.},
abstract = {IntroductionDual-tasking, the ability to perform two tasks
simultaneously or in close succession, has been associated
with increased fronto-parietal activity [1,2]. Furthermore,
functional and structural connectivity (FC and SC) display
unique features shown to be relevant to cognition and
performance; however, prediction studies show overall low
accuracies [3]. Accounting for inter-individual variability
in macroscopic brain organization may be informative for
elucidating such brain-behavior associations at the
individual level [4,5]. Here, we aimed to predict dual-task
performance from FC and SC in individualized task-specific
and whole-brain (WB) networks.MethodsWe obtained 92
individualized discrete and homologous functional regions
(Fig. 1A) from 71 healthy adults (33 females; 41 young and
30 older) by implementing an iterative cortical parcellation
approach based on functional surface-level connectivity
patterns [4,5] in combined resting-state and task-based fMRI
data. For each participant, Pearson correlations were
computed between the individualized parcels' time series to
derive FC estimates. The individualized parcels were
projected onto processed diffusion-weighted images and
normalized between-parcel streamline counts were calculated
to derive individualized SC estimates. Besides deriving WB
networks, connectomes were additionally obtained from
task-related parcels. These were defined through a
group-level general linear model of brain activity elicited
in a dual-task paradigm with spatially incongruent manual
responses to auditory stimuli [6,7]. The incongruent
dual-task condition recruited a large fronto-parietal
network [2], which overlapped with 25 out of 92 functional
regions (Fig. 1B). For comparison, we also computed FC and
SC between non-individualized parcels for both WB and
task-specific networks. To predict subject-level reaction
time in dual-tasking with incongruent responses (615.21 ±
118.38 ms), we implemented CPM, PLS, KRR and RFR as
incorporated in JuLearn [8] with a 5-times, 5-fold
cross-validation (CV) approach based on eight separate
feature spaces (i.e., FC vs. SC × task-specific vs. WB
networks × indiv. vs. non-indiv. networks). Prediction
performance was evaluated by the mean absolute percentage
error (MAPE), averaged across all models for all folds and
repetitions of the CV test set.ResultsThe best prediction of
dual-task performance was achieved by the individualized
task-specific and WB structural networks (MAPE = 15.13 ±
2.32 $\%,$ and MAPE = 15.34 ± $2.27\%,$ respectively), as
well as the individualized WB functional network (MAPE =
15.45 ± 3.18 $\%).$ This was followed by the
non-individualized WB functional network (MAPE = 15.73 ±
3.15 $\%).$ Predictability was lowest for the individualized
task-specific functional network (MAPE = 16.73 ±
3.00).ConclusionsWhile most brain-behavior prediction
studies use group-level brain atlases, our findings indicate
a slight improvement in prediction accuracy when accounting
for inter-individual variability in the brain's functional
organization, despite the overall low predictability and
small differences between feature spaces. Our results for
predicting dual-task performance align with an improvement
in predicting fluid intelligence from individualized
functional networks [4]. Counter to expectation, WB
functional networks outperformed the dual-task-specific
ones, which could be related to the relevance of global
brain organizational properties in brain-behavior
associations [9,10], whose individual nuances are better
captured by the WB functional connectome. Finally, we
suggest a replication study with a larger sample and the
better predictability from SC (vs. FC) calls for further
research integrating multi-modal brain features to explore
their unique and combined contribution to cognition [3].},
month = {Jul},
date = {2023-07-22},
organization = {Organization for Human Brain Mapping
(OHBM), Montreal (Canada), 22 Jul 2023
- 26 Jul 2023},
subtyp = {After Call},
cin = {INM-7},
cid = {I:(DE-Juel1)INM-7-20090406},
pnm = {5252 - Brain Dysfunction and Plasticity (POF4-525)},
pid = {G:(DE-HGF)POF4-5252},
typ = {PUB:(DE-HGF)24},
doi = {10.34734/FZJ-2023-03487},
url = {https://juser.fz-juelich.de/record/1014932},
}