Weak diffusion MRI mapping of brain structural connectivity associates with human cognitive ability
Poster No:
1182
Submission Type:
Abstract Submission
Authors:
Rong Wang1, Changsong Zhou2
Institutions:
1Xi'an Jiaotong University, Xi'an, shaanxi, 2Hong Kong Baptist University, Hong Kong, Hong Kong
First Author:
Co-Author:
Introduction:
Noninvasive tractography generates a significant degree of noise in weak, long-range streamline, but the sparse and backbone streamline appears to be stronger and more reliable and thus is supposed to be closer to the real fiber. With the general assumption of longer-range streamline having less contribution to tractography, serval post-filtering methods have been proposed to reduce the false-positive problems (e.g., commit2 and sift2) by compressing the streamlines, generating spare but reliable connectivity. Similar to the tractography field, network neuroscience also preferred the sparse and reliable connectivity by directly thresholding out weak connectivity to retain a low network density, as weak connectivity is believed to insignificantly affect the brain network topological measures and dynamics. Therefore, both the fields of tractography and network neuroscience have largely neglected the weak connectivity (usually less reliable or untraced). However, the sparse connectivity in human brain is strongly contradictory with the high density in mammalian animals' brain (e.g., 97% for 47 areas in mice) using reliable retrograde tracer injection. And the connectivity weight (i.e., number of axon projections) in animals spans several orders of magnitudes in these reliable data, contrary to the directly exclusion of weak connectivity and low network density. These evidences across species indicate the natural existence of weak fibers in the brain, and lead us to ask a critical, counter-intuitive open question whether the weak fiber in the human brain really does not contain any useful information and does not have impact on brain functions and dynamics.
Methods:
The probabilistic tractography algorithms were applied to map the streamlines which were further filtered by sift2 and commit2 methods. A machine learning model was designed to investigate whether weak connectivity with scaling parameter β is predictive of diverse cognitive abilities. We fused the tractography and filtering methods to extract more realistic SC by considering the trade-off between cognitive abilities predictions and network density. We constructed large-scale brain dynamic models to study how weak connectivity nonlinearly affects network dynamics. Then, we develop an eigenmode analysis method without thresholding that reflects the effect of weak connectivity across hierarchical levels of connectome activation modes. Finally, we investigate the heterogenous functions of weak connectivity to brain dynamics and cognitive abilities.
·Weak connectivity contributes to predictions of cognitive abilities
Results:
Accounting for the nonlinear functioning manner, weak connectivity is indispensable in predicting multiple cognitive abilities, which resists to noise permutation in weak connectivity. More realistic fusion SC was generated by balancing the trade-off between cognitive abilities predictions and network density, which with the nonlinear functional manner contributes to more realistic simulations of brain functional networks, and to better brain structure-function relationship. Weak connectivity displays an intersystem organization and supports the segregation-integration balance by enhancing network integration required by predicted cognitive abilities, but the structural mode analysis suggests that weak connectivity also promotes finer-scale local segregation. Multi-modal analyses find a category of weak connectivity that links low-order visual/motor to high-order limbic regions with negative gene coexpression (GC). This kind of weak connectivity has higher impact on network dynamics and predictions of cognitive abilities than that with positive GC.
Conclusions:
Weak connectivity functionally contributes to cognitive abilities by integrating different systems to maintain a balance between locally segregation and integration. The identified organization principles provide insights of the existence of two types of weak connectivity that potentially support forward and backward processing in the brain structural connectome.
Modeling and Analysis Methods:
Classification and Predictive Modeling
Connectivity (eg. functional, effective, structural) 1
Diffusion MRI Modeling and Analysis 2
fMRI Connectivity and Network Modeling
Keywords:
Cognition
FUNCTIONAL MRI
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
1|2Indicates the priority used for review