Date Published: March 5, 2019
Publisher: Public Library of Science
Author(s): Michael J. Paldino, Farahnaz Golriz, Wei Zhang, Zili D. Chu, Tavpritesh Sethi.
Architecture of the cerebral network has been shown to associate with IQ in children with epilepsy. However, subject-level prediction on this basis, a crucial step toward harnessing network analyses for the benefit of children with epilepsy, has yet to be achieved. We compared two network normalization strategies in terms of their ability to optimize subject-level inferences on the relationship between brain network architecture and brain function.
Patients with epilepsy and resting state fMRI were retrospectively identified. Brain network nodes were defined by anatomic parcellation, first in patient space (nodes defined for each patient) and again in template space (same nodes for all patients). Whole-brain weighted graphs were constructed according to pair-wise correlation of BOLD-signal time courses between nodes. The following metrics were then calculated: clustering coefficient, transitivity, modularity, path length, and global efficiency. Metrics computed on graphs in patient space were normalized to the same metric computed on a random network of identical size. A machine learning algorithm was used to predict patient IQ given access to only the network metrics.
Twenty-seven patients (8–18 years) comprised the final study group. All brain networks demonstrated expected small world properties. Accounting for intrinsic population heterogeneity had a significant effect on prediction accuracy. Specifically, transformation of all patients into a common standard space as well as normalization of metrics to those computed on a random network both substantially outperformed the use of non-normalized metrics.
Normalization contributed significantly to accurate subject-level prediction of cognitive function in children with epilepsy. These findings support the potential for quantitative network approaches to contribute clinically meaningful information in children with neurological disorders.
Pediatric epilepsy is a prototypical disorder of network dysfunction: even in the setting of a highly localized structural lesion, children with epilepsy demonstrate widespread alterations in cerebral cortical networks[1,2]. Whether established by genetic/developmental processes or by activity-dependent reorganization, and there is evidence to support a role for each, global network dysconnectivity undermines the brain’s capacity to support normal neuro-cognitive development[3–5]. Furthermore, the negative effects of epilepsy on intellectual function seem to be exaggerated in children, which may reflect the fact that developmental physiology is primed for cerebral growth and network reorganization. Regardless of origin, the impact of network dysfunction can be seen in the range and severity of cognitive failings exhibited by these children, often far beyond what would be expected based on the location and extent of their structural abnormalities. The ability to understand and predict the impact of global network dysconnectivity in an individual child with epilepsy would be of great value to the care of these patients. Current understanding points to the emergence of cognitive function from complex interactions occurring across large-scale brain networks that support both segregation into, as well as integration across, subspecialized systems. Non-invasive methodologies that capture the organization of the brain as a network of interacting elements, therefore, represent an appealing approach by which to study neurologic dysfunction in children with epilepsy.
We evaluated two strategies that aim to address the inter-individual variation in brain networks inherent to a clinical pediatric cohort. We specifically assessed the impact of these strategies on output metrics of global brain architecture in terms of their capacity to support the prediction of global intelligence in children with focal epilepsy. We report the following primary findings: 1. Normalization by either strategy significantly improved subject-level prediction of global intelligence; 2. Metric normalization in patient space outperformed the use of network registration into standard space under most conditions; and 3. Prediction improved across all conditions with increasing nodes in the network.
In conclusion, normalization contributed significantly to the prediction of individual intelligence in a cohort of children with focal epilepsy. Both of the tested normalization strategies significantly augmented prediction by the learning algorithm. However, under most conditions, normalization of metrics computed in patient space outperformed transformation of all patients into a standard space. These findings support the potential for network science to provide clinically meaningful markers of brain function in children with epilepsy.