Predicting dual-task performance from individualized functional and structural networks
Tue-Main hall - Z3-Poster 2-5917
Presented by: Lya K. Paas Oliveros
Dual-tasking is linked to fronto-parietal brain activity [1], and accounting for inter-individual variability in functional (FC) and structural connectivity (SC) between brain regions may be informative for individual performance [2-4]. Here, we aimed to predict dual-task performance from FC and SC in individualized task-specific (TS-N) and whole-brain (WB-N) networks.
In addition to standard, non-individualized connectivity measures, individualized FC and SC were derived from 41 young and 30 older adults by implementing an iterative cortical parcellation based on functional connectivity patterns [3,4]. Subject-level reaction time in dual-tasking was predicted via four regression models and a cross-validation approach based on eight separate feature spaces (FC vs. SC × TS-N vs. WB-N × indiv. vs. non-indiv. networks).
The best predictions were achieved with individualized structural TS-N and WB-N, followed by functional individualized and non-individualized WB-N. Predictability was lowest for the individualized functional TS-N.
This study revealed that considering inter-individual variability in functional brain organization via an individualized parcellation approach marginally improved dual-task performance prediction, aligning with previous studies showing limited prediction accuracy [5,6]. Additionally, WB-N outperformed TS-N, suggesting the significance of global brain organizational properties in brain-behavior associations [5,6].
[1] Paas Oliveros, LK, et al. (2023). Cereb Cortex, 33:10155–10180.
[2] Dhamala, E, et al. (2021). Hum Brain Map, 42:3102–3118.
[3] Wang, D, et al. (2015). Nat Neurosci, 18:1853–1860.
[4] Li, M, et al. (2019). PLOS Biol, 17:e2007032.
[5] Heckner, MK, et al. (2023). Cereb Cortex, 33:6495–6507.
[6] Pläschke, RN, et al. (2020). Cortex, 132:441–459.
In addition to standard, non-individualized connectivity measures, individualized FC and SC were derived from 41 young and 30 older adults by implementing an iterative cortical parcellation based on functional connectivity patterns [3,4]. Subject-level reaction time in dual-tasking was predicted via four regression models and a cross-validation approach based on eight separate feature spaces (FC vs. SC × TS-N vs. WB-N × indiv. vs. non-indiv. networks).
The best predictions were achieved with individualized structural TS-N and WB-N, followed by functional individualized and non-individualized WB-N. Predictability was lowest for the individualized functional TS-N.
This study revealed that considering inter-individual variability in functional brain organization via an individualized parcellation approach marginally improved dual-task performance prediction, aligning with previous studies showing limited prediction accuracy [5,6]. Additionally, WB-N outperformed TS-N, suggesting the significance of global brain organizational properties in brain-behavior associations [5,6].
[1] Paas Oliveros, LK, et al. (2023). Cereb Cortex, 33:10155–10180.
[2] Dhamala, E, et al. (2021). Hum Brain Map, 42:3102–3118.
[3] Wang, D, et al. (2015). Nat Neurosci, 18:1853–1860.
[4] Li, M, et al. (2019). PLOS Biol, 17:e2007032.
[5] Heckner, MK, et al. (2023). Cereb Cortex, 33:6495–6507.
[6] Pläschke, RN, et al. (2020). Cortex, 132:441–459.
Keywords: dual-tasking; inter-individual variability; functional connectivity; structural connectivity; machine learning