Technical Program

Paper Detail

Paper: PS-1B.8
Session: Poster Session 1B
Location: Symphony/Overture
Session Time: Thursday, September 6, 18:45 - 20:45
Presentation Time:Thursday, September 6, 18:45 - 20:45
Presentation: Poster
Publication: 2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania
Paper Title: Improving Corticostriatal Parcellation Through Multilevel Bagging with PyBASC
Manuscript:  Click here to view manuscript
DOI: https://doi.org/10.32470/CCN.2018.1141-0
Authors: Aki Nikolaidis, Child Mind Institute, United States; Joshua Vogelstein, Johns Hopkins University, United States; Pierre Bellec, University of Montreal, Canada; Michael Milham, Child Mind Institute, United States
Abstract: Understanding the functional organization the brain is a centrally important theme of human neuroscience. Ideally, these organizational maps uncover the underlying structure of the brain’s functional architecture, and group-level maps are accurate representations of the individuals in the sample. Using simulated fMRI data, we demonstrate that bagging improves the ability of clustering to uncover the data’s underlying structure. We show that the group-level maps become more correlated to the individual-level maps with more bootstrap aggregates, suggesting bagging improves the representativeness of the group-level solution. Using a test-retest dataset of 30 young adults, we confirm these findings. More specifically, we see bagging improves the test-retest correlation between cluster maps, and increases correlation between group-level and individual-level cluster maps, and these effects are robust to number of clusters and length of scan used. These results suggest bagging is an important method for increasing reliability and validity of functional parcellation approaches.